Numerical Methods for the Life Scientist

.
Heino Prinz

Numerical Methods
for the Life Scientist
Binding and Enzyme Kinetics Calculated
with GNU Octave and MATLAB

The source code of all programs is available

on extras.springer.com
Dr. Heino Prinz
Max-Planck-Institut für molekulare Physiologie
AG Biochemische Analytik
Otto-Hahn-Str. 11
44227 Dortmund
Germany
[email protected]

ISBN 978-3-642-20819-5 e-ISBN 978-3-642-20820-1

DOI 10.1007/978-3-642-20820-1
Springer Heidelberg Dordrecht London New York
Library of Congress Control Number: 2011933466

# Springer-Verlag Berlin Heidelberg 2011

This work is subject to copyright. All rights are reserved, whether the whole or part of the material is
concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting,
reproduction on microfilm or in any other way, and storage in data banks. Duplication of this publication
or parts thereof is permitted only under the provisions of the German Copyright Law of September 9,
1965, in its current version, and permission for use must always be obtained from Springer. Violations
are liable to prosecution under the German Copyright Law.
The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply,
even in the absence of a specific statement, that such names are exempt from the relevant protective
laws and regulations and therefore free for general use.

Printed on acid-free paper

Springer is part of Springer Science+Business Media (www.springer.com)

Preface

This textbook gives an introduction to numerical methods. They are key tools for
the calculation of reaction schemes and for the development of functional models
in life sciences. The most exciting of this is model development, but it probably
cannot be taught in a book. It is surprisingly personal. I will illustrate some aspects
together with an acknowledgment of the scientists who impressed me most.
Multiple theoretical approaches: I was fortunate to study physics at Bonn Univer-
sity at a time when Wolfgang Paul gave his main lectures on “Experimentalphysik”.
The lectures themselves were spectacular, but the emphasis was laid on the variety of
possible interpretations of the same experimental findings. For the topic of gravity,
for example, Wolfgang Paul included geocentric astronomy and concluded that one
cannot rule out angels carrying planets in retrograde loops around the earth, but that
the assumption of gravity is so much simpler. For an experimental physicist the
observation itself is the result. Theoretical models are only limited by the creativity of
the analyst. All are valid when they can be verified experimentally.
Courage and creativity: Friedrich Cramer was my PhD supervisor at the Max-
Planck Institute in Göttingen. He had published a structural model for tRNA based
on a map of reactive groups. When I entered his group, this model just had been
disproven, so that he was criticized for having published a preliminary model.
Model building is part of a scientific discourse, and one sometimes needs courage
to publish new ideas. Friedrich Cramer was a very creative, sometimes artistic, and
sometimes philosophical personality. He has strongly influenced me and many
other scientists in Göttingen.
Additional low-affinity binding: In the beginning of the 1980s, we had employed
a newly synthesized fluorescent cholinergic agonist for molecular studies of cho-
linergic excitation. For a detailed pharmacological characterization of this drug
Alfred Maelicke and I traveled to Göttingen, where we had convinced Bert
Sakmann and Erwin Neher to perform a series of electrophysiological studies at
different ligand concentrations. They confirmed that the ligand acted as a pure
agonist at low concentrations, but cautioned that it changed channel open times and
thereby acted as a local anesthetic at high concentrations on the same receptor. I as a

v
vi Preface

physicist who had calculated a model for cholinergic excitation found this particu-
larly disturbing, but they had observed such concentration dependency before. It is
well known that the physiological action of any compound depends on its dose,
but before this moment of truth I had thought that different target proteins were
responsible for the different interactions at different concentrations. Erwin Neher
and Bert Sakmann were the first to perform functional studies on one single
molecule. Independent of this, multiple interactions of ligands at the same target
molecule may be quite common as discussed in Sect. 4.8. Unfortunately, additional
low-affinity binding spoils the elegance of any mechanistic model. Additional sites
are at odds with many published crystal structures and rigid docking programs used
in drug design.
Open-minded observations and scientific communication: One cannot just sit
down and write a new theory without experimental evidence. One cannot even plan
for new functional models. Sometimes it just happens: One summer evening we
were sitting in a beer garden when Jörg Striessnig, an Innsbruck pharmacologist,
mentioned a discrepancy between a theoretical prediction and a common observa-
tion in the lab. He had observed that the addition of a second ligand to an existing
ligand–receptor complex resulted in dissociation rate constants, which increased
with the concentration of the second ligand. This observation corresponds to the
German saying “viel hilft viel” (the more you take, the more it helps), and therefore
“feels” all right, but, of course, it contradicts first-order dissociation kinetics. In the
end, we calculated the observation with the assumption of overlapping sites as
shown in Sect. 5.5. One really should talk science in relaxing environments.
Quantitative plausibility check: It is a good idea to present kinetic data to a
critical audience before publication. Some questions can be as basic as: “When you
have an equilibrium dissociation constant below nM, how can you have free pro-
tein?” The numerical values of rate constants imply scientific information which is
evident to an experienced scientist. I have often admired Roger Goody for his
overview.
Different scientific approaches: Scientists are individuals, who may have entirely
different approaches to the same question. These differences can lead to individual
animosities, but they also can lead to fruitful cooperation. Cooperation can be
encouraged when the person in charge has a natural human understanding coupled
to clear scientific goals. Herbert Waldmann is one scientist with such management
qualities and I am happy to work in his department.

Dortmund, Germany Heino Prinz

Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2 The Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1 Equations for First and Second Order Reactions . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Equilibrium Binding to Two Sites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3 Equilibrium Binding to Any Number of Sites . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.4 Writing Differential Equations for Two Site Binding . . . . . . . . . . . . . . . . 10
2.5 Writing Differential Equations for Any Reaction Scheme . . . . . . . . . . . . 12
2.6 Analytical and Numerical Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

3 Classical Analytical Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

3.1 Analytical Solutions for Equilibrium Binding . . . . . . . . . . . . . . . . . . . . . . . . 15
3.1.1 Direct Binding Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.1.2 Double Reciprocal Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.1.3 Comparison of Scatchard Plots and Double
Reciprocal Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.1.4 Lineweaver-Burk Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.1.5 Dose–Response Curves and Hill Coefficients . . . . . . . . . . . . . . . . . . 23
3.2 Analytical Solutions for Binding Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.2.1 Exponential Decay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.2.2 Half-Logarithmic Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2.3 Initial Velocity of Association Kinetics . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2.4 Pseudo First-Order Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.2.5 Multiple Exponential Fits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

4 Getting Started with Octave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

4.1 Installation Instructions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.2 Typing the First Commands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

vii
viii Contents

4.3 Writing a Program . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

4.4 Setting Up a Work Environment for Octave: Arrangement of Editor,
Terminal, and Plot Window . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.5 The MATLAB Environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.6 Arrays, Vectors, and Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.7 A Very Elementary Tutorial for GNU Octave and MATLAB . . . . . . . 39
4.8 Recommended Literature for Octave/MATLAB . . . . . . . . . . . . . . . . . . . . . . 40
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

5 Equilibrium Binding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.1 Solving Nonlinear Equations for Equilibrium Binding . . . . . . . . . . . . . . . 43
5.2 Equilibrium Binding to One Site (EQ1.m) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
5.3 Equilibrium Binding to Two Sites (EQ2.m) . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.4 Equilibrium Binding to Two Sites in the Presence
of Inhibitor (EQ3.m) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.5 Allosteric Interactions of Subunits (EQ4.m) . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.6 Allosteric Activators and Inhibitors (EQ5.m) . . . . . . . . . . . . . . . . . . . . . . . . . 62
5.7 Dose–Response Curves (EQ6.m) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
5.8 Multiple Allosteric Inhibition (EQ7.m) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

6 Binding Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
6.1 Solving Differential Equations in GNU Octave and MATLAB . . . . . . 71
6.2 Kinetics of Ligand Binding to One Site (kin1.m) . . . . . . . . . . . . . . . . . . . 73
6.3 Binding to One Site Followed by a Conformational Change
of the Receptor (kin2.m) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
6.4 Dissociation Kinetics: Chase with Inhibitor (kin3.m) . . . . . . . . . . . . . . . 81
6.5 Facilitated Dissociation (kin4.m) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
6.6 Sequential Binding and Obstructed Dissociation (kin5.m) . . . . . . . . . 87
6.7 Irreversible Inhibition (kin6.m) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95

7 Enzyme Kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
7.1 Progress Curves (enz1.m) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
7.2 Progress Curves for a Weak Substrate (enz1b.m) . . . . . . . . . . . . . . . . . . . 99
7.3 Michaelis–Menten Enzyme Kinetics (enz2.m) . . . . . . . . . . . . . . . . . . . . . 101
7.4 Lineweaver–Burk Plots from Progress Curves (enz3.m) . . . . . . . . . . . 102
7.5 Substrate Inhibition (enz4.m) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
7.6 Enzyme Inhibition with Reversible Inhibitors (Enz5.m) . . . . . . . . . . . 111
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

8 Fitting the Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119

8.1 Multi-parameter Fits and Correlation of Parameters (fit1.m) . . . . . . 119
8.1.1 The Sample Program fit1.m . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
Contents ix

8.1.2 Strategies to Fit Correlated Parameters . . . . . . . . . . . . . . . . . . . . . . . 123

8.2 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
8.3 Entering the Experimental Data (fit2.m) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
8.4 Fitting Substrate Inhibition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
8.4.1 Generating Sample Data (fit3.m) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
8.4.2 Calculating Steady State Equilibria (fit4.m) . . . . . . . . . . . . . . . . . 129
8.4.3 Global Fit of Progress Curves (fit5.m) . . . . . . . . . . . . . . . . . . . . . . . 136
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141

9 Appendix: Installing GNU Octave for Mac OS and Linux . . . . . . . . . . . 143

9.1 Installation Instructions for Mac OS X . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
9.2 Installation Instructions for Linux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
.
Chapter 1
Introduction

The analysis of quantitative data in biochemistry usually requires the calculation of

theoretical models. Their analytical solutions are functions given as explicit
formulas

y ¼ f(x)

The independent variable x may be the concentration of a substrate or the time,

and the computed value y may be the activity of an enzyme, or intrinsic fluores-
cence of a receptor, or any other parameter which can be computed from a theory.
Finding such an analytical function f(x) often has been a lengthy process. For
complex reaction schemes, it is easier to resort to numerical methods. These are
instructions sent to a computer to do the calculations.

Computer running
x ! a program with a ! y
set of instructions

Mathematically, these are also functions [1], even though they are not based on
explicit formulas. The set of computer instructions, or “algorithms” [2], assigns one
value y to each independent variable x. The output therefore is equivalent to any
other function y ¼ f(x). It may be understood with the “back box” concept [3],
whereby computer and its program are considered an unknown entity (black box)
which gives reproducible and accurate output y for a given input x. The calculations
described in this book are not concerned with details of the black box. They simply
use it. The resulting function is no formula, but is shown as a graph.
Only three types of numerical methods [4] are covered here, namely one method
to solve equations for binding equilibria, one method to calculate differential
equations for binding kinetics and one method to fit the data. Once the principles
are understood, it is easier to calculate complex reactions than to understand them
with complex simplifications.

H. Prinz, Numerical Methods for the Life Scientist, 1

DOI 10.1007/978-3-642-20820-1_1, # Springer-Verlag Berlin Heidelberg 2011
2 1 Introduction

Enzyme kinetics, binding kinetics and equilibrium binding have been studied for
more than 100 years. These topics are well presented in many modern textbooks
[5–11]. Their analysis is based on analytical solutions [12] and therefore is limited
to a given number of readily calculated reaction schemes. Numerical methods
overcome these limitations. They employ computers so that Chaps. 4–8 should be
read with an accompanying computer.
Programming is done in GNU Octave, a high-level language which is relatively
easy to learn and which includes simple graphical output. GNU Octave is available
for most computer platforms (Linux, windows, Mac, Sun, OS/2) and free to us all
[13]. It had originally been developed as companion software for a textbook
on reactor design [14]. Now it is continuously enhanced and refined by the GNU
community [15]. It is compatible to MATLAB®, a very popular commercial high-
level computer language [16]. MATLAB® is well supported by MathWorks™, its
proprietor, who also offers excellent training courses. Numerous lectures or
tutorials for MATLAB® and/or GNU Octave are available in many spoken
languages on the Internet. In some universities it belongs to the curriculum of the
mathematics departments. GNU Octave requires no financial investment and seems
to be the ideal software to get started. Readers with a MATLAB® license may
prefer the MATLAB® environment and should run the MATLAB® versions of the
programs.
GNU Octave is introduced step by step. The installation of the program is
described in the fourth chapter, and some basic commands are explained there.
From then on the language is applied to more than 50 practical examples of
increasing complexity. All sample programs, including the MATLAB® versions,
can be found in http://www.mpi-dortmund.mpg.de/misc/numericalmethods/. Each
sample program leads to a plot. Such a plot gives a relation of dependent and
independent values and is a graphical description of the numerical function x ! y.
An explicit formula is not required, and the plot illustrates the underlying mecha-
nism. In the end the readers and their computers should be able to calculate and fit
any reaction scheme, be it as complex as they like.
Exporting Octave programs to MATLAB® requires minor changes because the
Octave functions lsode and leasqr are similar but not identical to their
MATLAB® equivalents. The differences are explained when the functions are
used. The numerical methods introduced in this textbook are based on algorithms
available in most computer languages. When speed is an issue, one may wish to use
a high-level language with a fast compiler. One important example should be
mentioned; facsimile [17] is a dedicated program package based on the Fortran
Harwell Subroutine Library. It has been developed specifically for the calculation
of chemical reactions and is available under the GNU public license [13]. Unfortu-
nately, Fortran is not directly compatible with Octave or MATLAB.
The book is organized as follows: The second chapter covers the basics, namely
first- and second-order reactions. It explains how binding equilibria can be calcu-
lated from sets of ordinary equations, and how binding kinetics is calculated from
sets of differential equations. Analytical solutions of these equations are limited to
some examples, the most basics of which are summarized in Chap. 3. Chapter 4
1 Introduction 3

gives an introduction to GNU Octave. Chapter 5 covers equilibrium binding. The

first sample program, EQ1.m, is explained line by line, introducing all Octave
commands as they come along. It compares a numerical and an analytical solution
of simple reversible binding. From then on, the reaction schemes become more and
more complex, covering numeric solutions for more than one site, more than one
ligand and allosteric interactions. Chapter 6 deals with binding kinetics and its
calculations on the basis of differential equations. The sample programs focus on
the discrimination of molecular models. Enzyme kinetics is covered in Chap. 7,
where initial velocities and steady-state approximations are compared to progress
curves and numerical solutions of differential equations. Chapter 8 explains and
discusses the methods and limits of data fitting.
Different scientists may wish to use the book differently. A physicist may skim
through Chap. 2 to summarize the basics and then proceed to Chap. 3, which
explains common analytical methods in biochemical binding studies and enzyme
kinetics. He or she may skim through Chap. 4 before using the sample programs in
the following chapters as a tutorial for enzyme kinetics and Octave alike. A chemist
may skip Chap. 2, but should begin with Chap. 3, where simple reversible reactions
in biochemistry are summarized. A Biologist should begin with Chap. 4 and
focus on the principles of the computer language. Independent of the individual
approach, Chaps. 2 and 3 are written as reference for the sample programs provided
in Chaps. 5–8.
There are some practical aspects worth mentioning: Octave code and Octave file
names, line numbers, etc. will be set in Courier font throughout the book; the
units throughout the book are mM for the concentrations and seconds for the time; a
forward rate constant of 0.1 in program code thus translates to 0.1 mM1 s1, which
is equal to 105 M1 s1.
All sample programs are free as Open Access under the Creative Common
Attribution License [18]. Readers are encouraged to copy, distribute and modify
the programs as they like. They may be taken as seeds for further developments and
the readers are encouraged to share new programs at http://www.mpi-dortmund.
mpg.de/misc/numericalmethods/. Just send the new programs together with a short
description as an e-mail attachment to [email protected]. A
teaching course on numeric methods consists of a series of five lectures (2–3 h
each) covering Chaps. 4–8. It may be arranged with the author.

Acknowledgments This book has been written in the Max-Planck-Institute for molecular physi-
ology in Dortmund and owes to its stimulating scientific environment. Part of this book was
presented as a lecture to its IMPRS students http://www.imprs-cb.mpg.de/. They asked important
questions so that an improved course could be taught to students and faculty of the Suranaree
University of Technology http://www.sut.ac.th. This led to the development of a teaching course.
Alexander Fieroch introduced me to the free world of GNU software, and thus provided the
stimulus to write a textbook on numerical methods. He wrote the instructions to download GNU
Octave for Mac and Linux operating systems, which is provided in the appendix.
4 1 Introduction

References

1. Ruthing D (1984) Some definitions of the concept of function from Bernoulli, Joh. to
Bourbaki, N. Math Intel 6(4):72–77
2. Rogers H (1987) Theory of recursive functions and effective computability. The MIT Press,
Cambridge, MA
3. Beizer B (1995) Black-box testing: techniques for functional testing of software and systems.
Wiley, New York
4. Leader JJ (2004) Numerical analysis and scientific computation. Addison Wesley,
Reading, MA
5. Gutfreund H (1995) Kinetics for the life sciences. Receptors, transmitters and catalysts.
Cambridge University press, Cambridge
6. Segel IH (1993) Enzyme kinetics: behavior and analysis of rapid equilibrium and steady-state
enzyme systems. Wiley Classical Library, New York
7. Copeland RA (2005) Evaluation of enzyme inhibitors in drug discovery: a guide for medicinal
chemists and pharmacologists. Wiley-VCH, Weinheim
8. Copeland RA (2000) Enzymes: a practical introduction to structure, mechanism, and data
analysis. Wiley, New York
9. Purich DL (2010) Enzyme kinetics: catalysis & control: a reference of theory and best-
practice methods. Elsevier, London
10. Cook PF, Cleland WW (2007) Enzyme kinetics and mechanism. Garland Science, New York
11. Leskovac V (2003) Comprehensive enzyme kinetics, Kindle Edition. Amazon
12. Michaelis L (1912) Einf€ uhrung in die Mathematik f€ur Biologen und Chemiker. Springer,
Berlin
13. GNU general public license: http://www.gnu.org/licenses/gpl.html
14. Rawlings JB, Ekerdt JG (2002) Chemical reactor analysis and design fundamentals. Nob Hill
Publishing, Madison, WI
15. GNU Octave Homepage: http://www.gnu.org/software/octave/
16. MATLAB Homepage: http://www.mathworks.com/products/matlab/
17. Facsimile: http://www.facsim.org/ Powell MJD (1979) 25 years of theoretical physics
1954–1979: Chapter XVIII: numerical analysis. A special publication by Harwell Research
Laboratory of UKAEA
18. http://creativecommons.org/licenses/by/3.0/
Chapter 2
The Basics

Chemical reactions of the first and second order are the basis of all binding kinetics,
binding equilibria and enzyme kinetics. Kinetic experiments must be calculated
from a set of differential equations, whereas equilibrium binding studies can be
calculated from a set of ordinary equations. Chapter 2 provides guidelines to write
such equations for any reaction scheme. This is independent of the solution, be it
analytical or numerical.

2.1 Equations for First and Second Order Reactions

When a ligand L binds to a receptor R (2.1), the probability of interaction is

proportional to both concentrations. The rate of product formation d[LR]/dt there-
fore is the product of these concentrations times a constant, which is called the rate
constant k. Rate constants are denoted as lower case letters.

Second order association : L þ R ! LR (2.1)

Second order rates : d[LR]/dt ¼ k1  [L]  [R] (2.2)

d[L]/dt ¼ k1  [L]  [R] (2.3)

d[R]/dt ¼ k1  [L]  [R] (2.4)

These are differential equations. They have to be written for the concentration
changes of all components of the reaction, for L, R and LR. Note that we follow the
convention and write concentrations in square brackets. This convention cannot be
upheld in computer code, where a square bracket cannot be part of a variable name.
In octave and MATLAB, square brackets define a vector or matrix.
Once the complex LR has been formed, it may or may not dissociate. The
dissociation rate is proportional to the concentration of the complex.

H. Prinz, Numerical Methods for the Life Scientist, 5

DOI 10.1007/978-3-642-20820-1_2, # Springer-Verlag Berlin Heidelberg 2011
6 2 The Basics

First order dissociation : LR ! L þ R (2.5)

First order rates : d[LR]/dt ¼ k1  [LR] (2.6)

d[L]/dt ¼ k1  [LR] (2.7)

d[R]/dt ¼ k1  [LR] (2.8)

The negative sign in (2.6) indicates that the concentration of LR is decreased

upon the decay of the complex. Reaction schemes (2.1) and (2.5) together with the
basic reaction rates (2.2)–(2.4) and (2.6)–(2.8) contain all the physics required to
understand this textbook. Complex biochemical reactions result from bimolecular
association and monomolecular dissociation and combinations of these steps.
The above differential equations have been solved analytically, but let us use this
example to understand numerical solutions of differential equations: Mathemati-
cally, a differential quotient dx/dt is the limiting value of the difference quotient
Dx/Dt for infinitesimal small differences. Computer programs use a practical
version of this definition: Calculate the differences Dx of all concentrations for a
small time interval Dt, add these differences to the original values and do the
calculations for the next time interval. Repeat these steps until the time range of
interest is covered. That is all. For ordinary differential equations (they can be
written as dx/dt ¼ f(x, t), like all combinations of first and second order reactions)
octave provides the function lsode, a general procedure to solve them. The
function lsode will be covered in Chap. 6, but let us discuss the general principle
with the simple example of a reversible reaction:

k1
L þ R Ð LR (2.9)
k1

Reversible binding (2.9) leads to the formation of the complex LR and to a

corresponding decrease of L and R. The difference quotients are:

D[L]/Dt ¼ k1  [L]  [R] þ k1  [LR] (2.10)

D[R]/Dt ¼ k1  [L]  [R] þ k1  [LR] (2.11)

D[LR]/Dt ¼ k1  [L]  [R]  k1  [LR] (2.12)

When all rate constants and all initial concentrations are known, the concentra-
tion changes D[LR], D[L] and D[R] can be calculated from (2.10)–(2.12) for any
given time interval Dt.
Let us assume that the experiment simply consists of mixing receptor and ligand
to total concentrations R0 and L0, respectively. At time zero, the concentrations are
[R] ¼ R0, [L] ¼ L0, [LR] ¼ 0. After the small time interval Dt, these
concentrations are [R] ¼ R0 þ D[R], [L] ¼ L0 þ D[L] and [LR] ¼ 0 þ D[LR].
2.2 Equilibrium Binding to Two Sites 7

For the next time interval Dt these new concentrations have to be inserted in
(2.10)–(2.12) in order to compute the new concentration changes, and so forth.
These are very basic repetitive operations, ideally suited for a computer.
Equilibrium is reached for reversible reactions (2.9) when the dissociation rate
(1.6) is equal to association rate (2.2)

k1  [L]  [R] ¼ k1  [LR] (2.13)

This leads to (2.14) with the equilibrium dissociation constant KD1. Equilibrium
constants K1 ¼ k1/k1 and equilibrium dissociation constants KD1 ¼ k1/k1 are
denoted with uppercase letters. In life science and in this textbook, equilibrium
dissociation constants KD are used rather than equilibrium constants K. They
correspond to the ligand concentration where half of the binding sites for this KD
are occupied.

KD 1 ¼ k1 =k1 ¼ [L]  [R]/[LR] (2.14)

[LR] ¼ [L]  [R]/KD 1 (2.15)

Bound ligand and free ligand concentrations must add up to the total ligand
concentration L0. The same must hold for all components of bound and free
receptors which add up to the total receptor concentration R0:

R0 ¼ [R] þ [L]  [R]/KD (2.16)

L0 ¼ [L] þ [L]  [R]=KD (2.17)

Equations (2.16) and (2.17) are two equations with the two unknowns [L] and
[R]. They follow the pattern: “The total concentration of each molecule must be the
sum of free and bound concentrations”. This pattern leads to n equations of n
unknowns. Even the most complex equilibria can be calculated from these
equations. Analytical solutions for such general equations involving numerous
complexes and numerous ligands may be difficult or impossible to find, but
numerical solutions are relatively easy with the help of a suitable algorithm.
Again, as has been discussed for differential equations, the main task for the
scientist lies in writing the equations.

2.2 Equilibrium Binding to Two Sites

The binding of ligands to two sites of a receptor is often described with reaction
scheme (2.18), where a ligand may bind to two sites. Only one singly bound
complex LR and one doubly bound L2R are considered in this scheme. The scheme
8 2 The Basics

can be extended to incorporate more binding sites, always with only one complex
and one equilibrium dissociation constant for each sequential binding step

k1 k2
2L þ R Ð L þ LR Ð L2 R (2.18)
k1 k2

Scheme (2.18) is ideally suited for the analysis of equilibrium binding studies,
since it involves a minimal number of constants. It is, however, not a plausible
scheme. Generally, one would expect the ligand to have a choice for binding to one
or to the other site, before the second ligand binds to the non-occupied site. Such a
mechanism of accessible sites is shown in scheme (2.19) where the first ligand may
bind to the “left” (LR) or “right” (RL) site, before a second ligand leads to the
formation of the fully occupied receptor LRL.

ð2:19Þ

Reaction scheme (2.19) has sometimes been called “diamond model” because of
its shape. It illustrates the most general model for two site binding. Sites are
considered to be independent when the ligand bound to one site has the same
affinity independent of the occupation of the other site. This statement translates
to KD1 ¼ KD4 and KD2 ¼ KD3. The sites are equivalent when their affinity is the
same, i.e. KD1 ¼ KD2. For equivalent sites, binding is regarded to be cooperative,
when KD1 > KD3, so that the second ligand binds with a higher affinity. For
KD1 < KD3 and equivalent sites, the binding mechanism is called anticooperative.
One should note that reaction scheme (2.19) shows a coupled equilibrium and
that the affinity of the ternary complex LRL must be independent of the path by
which it was formed.

Upperpath : [LR] ¼ [L]  [R]/KD 1 [LRL] ¼ [LR]  [L=KD 3 (2.20)

Lowerpath : [RL] ¼ [L]  [R]/KD 2 [LRL] ¼ [RL]  [L]/KD 4 (2.21)

If we combine all four equations, we get

[L]  [L]  ½R=ðKD 1  KD 3Þ ¼ ½L  ½L  ½R=ðKD 2  KD 4Þ (2.22)

And thus

KD 1  KD 3 ¼ KD 2  KD 4 (2.23)
2.3 Equilibrium Binding to Any Number of Sites 9

One equilibrium dissociation constant of a coupled equilibrium can therefore be

calculated from the other ones. This principle holds for all coupled equilibria, since
the affinity of a complex must be independent of the path by which it was formed.
The concentrations of the complexes LR, RL and LRL, written in (2.20)–(2.22) are
functions of free concentrations [L] and [R]. The total concentrations R0 and L0
must therefore be equal to

R0 ¼ [R] þ [L]  [R]/KD 1 þ [L]  [R]/KD 2 þ [L]  [L]  [R]=ðKD 1  KD 3Þ (2.24)

L0 ¼ [L] þ [L]  [R]/KD 1 þ [L]  [R]/KD 2 þ 2  [L]  [L]  [R]/(KD 1  KD 3Þ (2.25)

Note that the complex LRL contains two bound ligand molecules. This is taken
into account by the factor 2 in (2.25). For reaction scheme (2.18), the corresponding
sets of equations are given in (2.26) and (2.27):

R0 ¼ [R] þ [L]  [R]/KD 1 þ [L]  [L]  [R]/(KD 1  KD 2Þ (2.26)

L0 ¼ ½L þ ½L  [R]=KD 1 þ 2  [L]  [L]  [R]=ðKD 1  KD 2Þ (2.27)

Note that KD2 in reaction scheme (2.18) indicates the occupation of the second
ligand to form the fully saturated complex LRL, whereas KD2 in reaction scheme
(2.19) denotes the binding of one ligand to an independent second site. Equations
(2.24) and (2.25) can be re-written as (2.28) and (2.29).

R0 ¼ [R] þ [L]  [R]  ð1=KD 1 þ 1=KD 2Þ þ [L]  [L]  [R]/(KD 1  KD 3Þ (2.28)

L0 = [L] þ ½L  [R]  ð1=KD 1 þ 1=KD 2Þ þ 2  [L]  [L]  [R]/(KD 1  KD 3Þ (2.29)

They are identical when 1/KD1 for scheme (2.18) equals (1/KD1 þ 1/KD2) from
scheme (2.19) and KD1 · KD2 from scheme (2.18) equals KD1 · KD3 from scheme
(2.19).
Equilibrium binding studies therefore cannot distinguish between the sequential
scheme (2.18) and a more plausible scheme of accessible sites (2.19). It will be
shown in Sect. 6.5 that a sequential binding mechanism (2.18) can be identified with
a kinetic “chase” experiment. But only a kinetic experiment with two different
ligands allows the distinction between first and second bound ligand.

2.3 Equilibrium Binding to Any Number of Sites

These conclusions can be generalized: Equilibrium binding studies can only deter-
mine one effective affinity for each occupation of sites, no matter how many
individual complexes can be identified for each monoliganded, diliganded,
10 2 The Basics

triliganded, etc. occupation. If we know the intrinsic affinities of each individual

binding site at each step, we can compute the simplified affinities of a sequential
scheme. This reasoning does not work in the opposite direction: Since equilibrium
binding studies can only determine one effective equilibrium dissociation constant
for the different occupation of sites, intrinsic equilibrium dissociation constants for
the different sites cannot be deduced without further assumptions.
Only if we assume that affinities of all n individual binding sites are equivalent,
can intrinsic equilibrium dissociation constants KDiintrinsic for the ith steps be
calculated from the effective (sequential) equilibrium dissociation constants (2.30).

KD iintrinsic ¼ ((n  i þ 1)/i)  KD isequential (2.30)

As an example, let us assume two step sequential binding sites with KD1sequential ¼
KD2sequential ¼ 10 mM according to reaction scheme (2.18). These equilibrium disso-
ciation constants of scheme (2.18) translate to KD1intrinsic (¼KD2intrinsic) ¼ 20 mM and
KD3intrinsic (¼KD4intrinsic) ¼ 5 mM of reaction scheme (2.19). Therefore, identical
equilibrium dissociation constants for each step of a sequential mechanism (like
(2.18)) translate into cooperative binding when intrinsic equilibrium dissociation
constants of accessible sites are computed.
Calculating equilibrium binding curves from any given reaction scheme is quite
straight-forward. One simply writes one equation (2.31) for each molecule of the
reaction

Total concentration ¼ free concentration þ bound concentration (2.31)

This corresponds to (2.24) and (2.25). The bound concentration is the sum of all
molecules bound to all complexes. The set of (2.31) can be solved numerically.
These equations need not be linear, and there may be more than one solution.
A reasonable initial estimate ensures that the right solution is found. This is
discussed in Chap. 5.

2.4 Writing Differential Equations for Two Site Binding

Differential equations can readily be solved with numerical methods, but they have
to be set up first. For example, analyzing reaction scheme (2.18) involves the
calculation of four concentrations, namely [R], [L], [LR], and [L2R]. The four
differential equations for the four concentration changes are given in (2.32)–(2.35)

d[R]/dt ¼ k1  [R]  [L] þ k1  [LR] (2.32)

d[L]/dt ¼ k1  [R]  [L] þ k1  [LR]  k2  [LR]  [L] þ k2  ½L2 R] (2.33)
2.4 Writing Differential Equations for Two Site Binding 11

d[LR]/dt ¼ k1  [R]  [L]  k1  [LR]  k2  [LR]  [L] þ k2  ½L2 R] (2.34)

d[L2 R]/dt ¼ k2  [LR]  [L]  k2  ½L2 R] (2.35)

Note the algebraic signs. Since all concentrations and rate constants are positive,
a decrease in concentration, that is a negative rate, always is indicated by a negative
sign. Note (2.34), where LR is decreased by the dissociation part of reaction 1 and
the association part of reaction 2. Any second order reaction involves three
components of the reactions. Therefore, reaction 1 with k1 and k1 affects the
concentration changes of R, L and LR in (2.32), (2.33) and (2.34). Likewise,
reaction 2 with k2 and k2 involves the components L, LR and L2R and thus affects
(2.33), (2.34) and (2.35). For more complex schemes, such a check that a given rate
constant must be involved in three (or two, for reversible first order) reactions helps
in debugging a program. One can use the “find” function included in any text editor
to make sure that each rate constant of a reversible reaction of the second order
appears three times.
Differential equations for reaction scheme (2.19) require the calculation of five
concentrations, namely [R], [L], [LR], [RL], and [LRL]. This is shown in equations
(2.36)–(2.40):

d[R]/dt ¼ k1  [R]  [L] þ k1  [LR]  k2  [R]  [L] þ k2  [RL] (2.36)

d[L]/dt ¼  k1  [R]  [L] þ k1  [LR]  k3  [LR]  [L] þ k3  [LR]

(2.37)
 k2  [R]  [L] þ k2  [LR]  k4  [RL]  [L] þ k4  [LRL]

d½LR/dt ¼ k1  ½R  ½L  k1  ½LR  k3  ½LR  ½L þ k3  ½LRL (2.38)

d[RL]/dt ¼ k2  [R]  [L]  k2  [RL]  k4  [LR]  [L] þ k4  [LRL] (2.39)

d[LRL]/dt ¼ k3  [LR]  [L]  k3  [LRL] þ k4  [LR]  [L]  k4  [LRL] (2.40)

The more complex a reaction scheme becomes, the more and longer differential
equations have to be computed. Each reversible reaction corresponds to two
products of rate constant and component of the reactions. Note that the free ligand
L is involved in all four reactions, so that there are eight products in (2.37). There
are two reactions involving the free receptor R in reaction scheme (2.19), so that
there are four products in (2.36). Likewise, two pathways for the dissociation of the
complex LRL are reflected in four products in (2.40).
One important restriction must be considered: Reaction scheme (2.19) contains
a closed loop which corresponds to coupled equilibria calculated in (2.23).
This reduces the number of independent rate constants from 8 (4 reversible
reactions) to 7. Equation (2.23), written for rate constants instead of equilibrium
constants, translates into:
12 2 The Basics

k1  k3  k2  k4 ¼ k2  k4  k1  k3 (2.41)

Again (2.41) ensures consistency, just like (2.23). It may be interpreted as a rule
for circular reactions: The product of rate constants in one (clockwise) direction of
reaction scheme (2.19) must be the same as the product of rate constants in the other
(counterclockwise) direction. One of the rate constants in a closed loop therefore
can be calculated from the other ones.

2.5 Writing Differential Equations for Any Reaction Scheme

Writing differential equations for a complex reaction scheme may look compli-
cated, but it does not require much fantasy. First, one counts all complexes and free
ligands. Then one has to write one, and possibly a long one, differential equation for
the concentration change of each of these components. For reversible reactions in a
fixed volume in solution, these concentration changes can only be expected from
second or first order reactions. Each second order reaction has to be considered in
the concentration changes of all three components of the reaction. There may be
first order conformational changes which involve only two components (one for
each conformation) or first order decay which only involves one component,
provided the product is irrelevant and the decay is not reversible. Closed circles
in reaction schemes have to fulfill the criterion of (2.41), namely that the products
of all rate constants in one direction must be the same as the products of rate
constants in the other direction.
That is all. But when differential equations are written as part of a program code,
no typing error is allowed. It helps to use a “find” function in a program editor and
look for all the rate constants individually. Any rate constant involved in a bimo-
lecular reversible reaction must appear thrice. Whenever a given concentration
appears in the differential equation of this concentration, the algebraic sign in front
of the accompanying rate constant must be negative. For reversible reactions, the
number of products in each differential equation must be even. Those little controls
may help.

2.6 Analytical and Numerical Solutions

Only for the simplest cases, the sets of equations described above can be solved
analytically. But when an analytical solution is found, it is precise and reliable for
all feasible concentrations. Finding analytical solutions needs a lot of effort, but
calculating them can be done from one formula with simple spread sheets or pocket
calculators. Some of the most important analytical solutions are covered in Chap. 3.
2.6 Analytical and Numerical Solutions 13

Numerical methods are different. They are well established, but they depend on
a computer and a program to run them with. The algorithms are based on
approximations. The results are the same within reasonable errors, but they are
not identical to analytical formulas. Figure 5.2 illustrates this with one example. All
calculations done by computers are limited by the precision of the stored variables.
When small differences of large numbers approach the precision of those large
numbers, they become unreliable. Octave and MATLAB typically issue warnings
when the internal precision is not sufficient. In general, whenever stochastic results
are computed with numerical methods, one should repeat the calculations with
other parameters (concentrations or rate constants). Use a computer, but keep
checking it and do not develop unconfined trust!
.
Chapter 3
Classical Analytical Solutions

Reversible binding to one site can be calculated with analytical solutions. On the
basis of these formulas, data can be transformed to appear linear. Straight lines of
equilibrium-binding studies in double reciprocal plots or of enzyme kinetics in
Lineweaver-Burk plots or of dissociation kinetics in half logarithmic plots indicate
simple mechanisms. Deviations from these are discussed in detail as cooperative or
independent sites. Logistic functions commonly used to calculate dose–response
curves, only correspond to a binding mechanism when the Hill coefficient is one. At
the end of this chapter, the reader should be able to identify simple biochemical
reactions.

3.1 Analytical Solutions for Equilibrium Binding

Simple reversible binding (scheme (2.9)) under equilibrium conditions may be

calculated from (2.15) to (2.17). Computing [R] from (2.16) and inserting it into
(2.15) results in the well-known hyperbolic binding function:

[LR] ¼ [L]  R0/([L] þ KD Þ (3.1)

3.1.1 Direct Binding Curves

Figure 3.1 shows a binding curve corresponding to (3.1) with a receptor concentra-
tion of 100 mM, a KD value of 10 mM and a maximum free ligand concentration of
100 mM. It is characterized by a linear, maximal concentration increase at low free
ligand concentration, followed by a steady decrease in its slope, leading to an
asymptotic approach to its maximal value at high ligand concentrations.
Almost all figures in this textbook show the calculations with Octave programs,
and the names of the program files (like ana1.m) are indicated in brackets after

H. Prinz, Numerical Methods for the Life Scientist, 15

DOI 10.1007/978-3-642-20820-1_3, # Springer-Verlag Berlin Heidelberg 2011
16 3 Classical Analytical Solutions

the title. The program files are listed as source code and are supplied in the
supplementary material. They are not the topic of Chap. 3, but may serve later as
independent exercises.
The maximal bound ligand is equal to the total receptor concentration R0 ¼ 100
mM, and is indicated as a solid line in Fig. 3.1. The concentration of free ligand
which leads to 50% saturation of receptor (50 mM bound ligand) is equal to the
equilibrium dissociation constant KD. This is also indicated as a solid line in
Fig. 3.1.
Figure 3.1 shows the calculation with a receptor concentration ten times the
equilibrium dissociation constant. Usually, binding curves are measured with recep-
tor concentrations around the KD value. This is not relevant for the shape of the curve
when bound ligand is plotted versus the free ligand concentration (Fig. 3.1 and (+)
in Fig. 3.2). When bound ligand is plotted versus total ligand concentration ((o)
in Fig. 3.2), the curves differ, depending on receptor concentration and affinity. For
extremely high affinities (low KD value) or for irreversible reactions, the ligand binds
quantitatively until saturation is reached. This yields a straight line with a sharp bend,
the solid line in Fig. 3.2. The bend marks the total receptor concentration R0.
KD values can be determined directly from 50% binding only when binding is
plotted versus the free ligand concentration (Fig. 3.1). When plotted versus total
concentration (Fig. 3.2), the curves may differ considerably, depending on the
receptor concentration. In most cases the receptor concentration is in the order of

Equilibrium Binding curve (ana1.m)

100
Concentration of [LR] (bound ligand)

80

60

40

20

0
0 20 40 60 80 100
Concentration of [L] (free ligand)

Fig. 3.1 Equilibrium-binding curve (bound ligand vs. free ligand), calculated with R0 ¼ 100 mM
and KD ¼ 10 mM. The maximal bound ligand R0 and half-maximal bound ligand are indicated by
solid lines, giving the asymptote and the KD values, respectively
3.1 Analytical Solutions for Equilibrium Binding 17

Equilibrium Binding curve vs. total and free ligand concentration (ana2.m)

Concentration of [LR] (bound ligand) 100

80

60

40

20

0
0 50 100 150 200
Concentration of L0 (o) or [L] (+) (total or free ligand)

Fig. 3.2 Equilibrium-binding curve vs. total and free ligand concentration. R0 ¼ 100 mM,
KD ¼ 10 mM. Bound ligand [LR] is plotted vs. free (+) and total (o) ligand concentration. Analytical
solutions are calculated with ana2.m. The solid line corresponds to the limiting case of extremely
high affinity or irreversible binding, where all ligands are bound and no free ligand is available until
the receptor is saturated. The sharp bend therefore is observed at L0 ¼ R0 ¼ 100 mM

magnitude lower than the ligand concentration. For these common cases, bound
ligand is only a tiny fraction of the total ligand concentration so that the difference
between total and free ligand concentration can be disregarded. The concentrations
of pharmacological receptors are much more than thousand times lower than the
administered drugs so that 50% occupation of these receptors is expected when the
total drug concentration (at the target tissue) equals to the KD value. If a physiolog-
ical effect is caused only by the binding of a drug to the receptor molecule [1], then
half-maximal binding to this receptor will lead to a half-maximal effect. The
effective drug concentration for 50% activity (EC50) must therefore be equal the
equilibrium dissociation constant KD of the drug to the receptor.

3.1.2 Double Reciprocal Plots

If one takes the inverse values of both sides of (3.1), it becomes:

1/[LR] ¼ (1/[L])  ðKD /R0Þ þ 1/R0 (3.2)

18 3 Classical Analytical Solutions

Double reciprocal Plot (ana3.m)

0.04

0.035

0.03
1 / [LR] (1/ bound ligand)

0.025

0.02

0.015

0.01

0.005

0
-0.1 -0.05 0 0.05 0.1 0.15 0.2 0.25 0.3
1 / [L] (+) or 1 / L0 (o)

Fig. 3.3 Double reciprocal Plot with R0 ¼ 100 mM, KD ¼ 10 mM (same data as shown in
Fig. 3.2). 1/[LR] is plotted vs. 1/free (+) and 1/total (o) ligand concentration. The theoretical
straight line is extrapolated to (0, 1/KD)

Therefore, when 1/[LR] is plotted versus 1/[L], the resulting curve should be a
straight line with a slope of KD/R0 and the intersection of 1/R0 at the y-axis. If the
line is extrapolated toward negative values, the intersection with the x-axis will be
at 1/KD. This is shown in Fig. 3.3.
Indeed the double reciprocal plot of bound and free ligand (indicated by the
symbol (+)) is linear. Many functions appear linear, or almost linear, in a double
reciprocal plot (compare Figs. 3.4 and 3.5). A double reciprocal plot of bound
ligand versus total ligand ((o) in Fig. 3.2) also appears linear. If this curve would
inadvertently be fitted to a straight line, the resulting KD values and receptor
concentrations might be far too high.

3.1.3 Comparison of Scatchard Plots and Double Reciprocal Plots

Double reciprocal plots (Fig. 3.3) sometimes appear linear, even when the plotted
function is not a simple binding curve (o). Another transformation of (3.1) is shown
in (3.3)

[LR]/[L] ¼ [LR]/KD þ R0=KD (3.3)

3.1 Analytical Solutions for Equilibrium Binding 19

Scatchard Plots (ana4.m)

0.05
one binding site
cooperative binding
2 independent sites
0.04
Bound / Free ligand concentration

0.03

0.02

0.01

0
0 0.5 1 1.5 2
Concentration of bound ligand

Fig. 3.4 Scatchard Plots. R0 ¼ 1 mM. (+) Binding to one site with KD ¼ 10 mM. (o) cooperative
binding to two equivalent sites (scheme 1.19, KD2 ¼ KD1), KD1 ¼ 100 mM, KD3 ¼ 10 mM. (x)
Two independent binding sites [scheme (2.19), KD3 ¼ KD2] with KD1 ¼ 10 mM, KD2 ¼ 100 mM.
The theoretical curves plotted are shown as solid lines. A random noise of 5% was assumed for
both free and bound ligand. Data points were calculated from logarithmic distributions of ligand
concentrations

Plotting bound ligand divided by free ligand versus the bound ligand concentra-
tion will give a straight line with a negative slope 1/KD. This is called a Scatchard
plot [2–5] and shown in Fig. 3.4. The corresponding double reciprocal plot is shown
in Fig. 3.5.
Scatchard plots are more sensitive or responsive than double reciprocal plots.
Higher sensitivity also means that experimental errors influence the shape of the
curve to a larger extent. This becomes obvious when theoretical curves (solid lines
shown in Figs. 3.4 and 3.5) are supplemented with “experimental” data
corresponding to a 5% random noise.
The data points shown in Figs. 3.4 and 3.5 were calculated from ligand
concentrations distributed in a logarithmic scale. Logarithmic distributions help
to get more significant information with the same number of data points. This can
be seen when a double reciprocal plot with linear distribution of data points
(Fig. 3.3) is compared with the same type of plots and a logarithmic distribution
(Fig. 3.5). Linear distributions plotted in a double reciprocal scale simply do not
have enough data points at low ligand concentrations. Experimental noise of these
sparsely distributed points influences the shape of the curve disproportionally.
Transformed data such as generated from double reciprocal or Scatchard plots
20 3 Classical Analytical Solutions

Double reciprocal plot (ana4.m)

3
one binding site
cooperative binding
2.5 2 independent sites
1 / bound ligand concentration

1.5

0.5

0
0 0.02 0.04 0.06 0.08 0.1
1 / free ligand concentration

Fig. 3.5 Double reciprocal plot (same data as shown in Fig. 3.4) R0 ¼ 1 mM. (+) Binding to one
site with KD ¼ 10 mM. (o) Cooperative binding to two sites, KD1 ¼ 100 mM, KD2 ¼ 10 mM. (x)
Two independent binding sites with KD1 ¼ 10 mM, KD2 ¼ 100 mM. The theoretical curves
plotted are shown as solid lines. A random noise of 5% was assumed for both free and bound
ligand. Data points were calculated from logarithmic distributions of ligand concentrations

generally should not be used for data fitting because the experimental errors are not
proportional to the transformed data values. Fits to the original data are more
reliable. They require nonlinear fitting routines which are covered in Chap. 8.
A simple binding curve (3.1) gives a straight line in a Scatchard plot ((+) in
Fig. 3.4). The interpolation to the x-axis gives the concentration of maximal bound
ligand. For one binding site on the receptor molecule, this is identical to the receptor
concentration R0. If there are two binding sites on the receptor, this value increases
by the factor 2. These sites may have different affinities for the same ligand. If one
binding site is characterized by an equilibrium dissociation constant of 10 mM,
and the other by 100 mM, the data show a marked curvature (x) in the Scatchard plot
(Fig. 3.4), whereas the same data almost appear linear in a double reciprocal
plot (Fig. 3.5). This is the reason why Scatchard plots have been used to distinguish
between cooperative (o) and noncooperative (x) binding. It should be mentioned
that a Scatchard Plot is the same as an Eadie-Hofstee plot when y- and x-axes are
exchanged.
The comparison of Figs. 3.4 and 3.5 shows that Scatchard plots are superior to
detect systematic deviations from simple binding equilibrium. Double reciprocal
plots always have a tendency to appear linear.
3.1 Analytical Solutions for Equilibrium Binding 21

3.1.4 Lineweaver-Burk Plots

Enzymes are biopolymers which catalyze a biochemical reaction. In the simplest

case, an enzyme E binds a substrate S and forms the complex ES. The product P is
formed in the enzyme. Both the product formation and dissociation are first-order
processes. They may be combined to one first-order catalytic rate constant kcat so
that reaction scheme (3.4) may be used.

k1 kcat
S þ E Ð ES 0 E þ P (3.4)
k1

When the dissociation rate constant k1 is much larger than kcat, only a fraction of
ES dissociates to E + P. When substrate is in large excess over E, only a tiny fraction
of S can bind and only a small part of this can be catalyzed to form P. In this case, the
initial equilibrium of S + E and ES is maintained for quite a while until a significant
amount of S has been used up by the catalysis. This initial part of such a reaction is
called “steady-state equilibrium” [6]. It can be computed just like any other equilib-
rium [7], with an equilibrium dissociation constant Km ¼ (k1 + kcat)/k1. Km is
called the Michaelis constant and steady-state enzyme kinetics of reaction scheme
(3.4) are referred to as “Michaelis-Menten kinetics” [5, 8–13].
The rate of product formation follows from first-order dissociation (2.6)

v ¼ d[P]/dt ¼ kcat  [ES] (3.5)

The enzyme concentration E0 usually is more than a factor of hundred smaller

than the substrate concentration so that in the beginning of the reaction the free
ligand concentration does not differ significantly from the total ligand concentration
([L] ¼ L0). This gives a steady-state equilibrium, and the concentration of bound
substrate [ES] can be computed from (3.1). Substituting L with S0, R0 with E0, and
KD with Km leads to

v0 ¼ d[P]/dt ¼ kcat  S0  E0/(S0 þ Km Þ (3.6)

The maximal velocity is

vmax ¼ kcat  E0 (3.7)

and (3.6) becomes

v0 ¼ d[P]/dt ¼ vmax  S0 =ðS0 þ Km Þ (3.8)

Equation (3.8) is called Michaelis-Menten equation [5, 8–13]. It is basically the

same as (3.1), but it calculates the initial velocity of product formation (or substrate
depletion) rather than ligand binding. The Michaelis constant is equal to the
22 3 Classical Analytical Solutions

equilibrium dissociation constant of substrate binding when kcat << k1. Other
cases of steady-state approximations will not be discussed in this chapter, because
numerical methods allow correct calculations. The program enz2.m (Fig. 7.3)
calculates reaction scheme (3.4) without any restrictions imposed on the selection
of rate constants.
When the initial velocity is plotted versus the substrate concentration, the
resulting curve likewise is very similar to that shown in Figure 3.1. It is an
equilibrium-binding curve attributed to the steady-state equilibrium of enzyme
kinetics. Sometimes this type of plot itself is referred to as “enzyme kinetics” or
“Michaelis-Menten kinetics.” Just as shown in (3.2) for direct binding curves,
taking the inverse of (3.8) will give

1=v0 ¼ ðKm =vmax Þ  1=S0 þ 1=vmax (3.9)

Therefore, when 1/v0 is plotted versus 1/S0, the result will be a straight line with
a slope of Km/vmax and an intersection 1/vmax at the y-axis. The double reciprocal
plot for enzyme kinetics is called Lineweaver-Burk plot [14] (Fig. 3.6).

Lineweaver Burk Plot (ana5.m)

80

70

60
1 / [v0] (1/ initial velocity)

50

40

30

20

10

0
-0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
1 / S0

Fig. 3.6 Lineweaver-Burk plot. Double reciprocal plot of initial velocity and total substrate
concentration. Initial velocity is calculated from (3.8) with KM ¼ 10 mM, E0 ¼ 10 nM, kcat ¼ 10 s1.
Intersection with the x-axis yields 1/KM, the intersection with the y-axis 1/vmax. The theoretical
curve plotted is shown as a solid line. A random noise of 10% was assumed for the initial velocity.
Data points were calculated from a twofold dilution series of the substrate concentration beginning
with 100 mM
3.1 Analytical Solutions for Equilibrium Binding 23

Figure 3.6 shows the calculation from substrate concentrations derived from a
twofold dilution series. Section 8.2 explains why such a series ensures a minimal
experimental error. Comparing Figs. 3.6 and 3.3 shows that such a dilution series is
much more effective in covering a more significant data range. Only seven data
points were required for the calculation shown in Fig. 3.6.

3.1.5 Dose–Response Curves and Hill Coefficients

When drug binding to a receptor [1] leads to a physiological response, the intensity
of this response should follow the same curve as the binding curve (3.1). Such a
dose–response curve usually is plotted in a logarithmic scale, where a hyperbolic
function (Fig. 3.1) will look sigmoid ((+) in Fig. 3.7). In many cases, experimental
dose–response curves do not follow (3.1). The maximal slope of the sigmoid
dose–response curve often is larger than that expected from binding to one site.
In this case, scheme (3.10), which had been used in 1910 for the binding of oxygen
to hemoglobin [15], often is used as an approximation.

nL þ R Ð Ln R (3.10)

Logistic dose-response curves (ana6.m)

100
Binding curve, nH = 1
Logistic curve, nH = 2
Logistic curve, nH = 4
Response (% of maximal response)

80

60

40

20

0
0 0.5 1 1.5 2
log (ligand concentration)

Fig. 3.7 Logistic dose–response curves. The binding curve (+) was calculated from (3.1), the
logistic curves from (3.12) with n ¼ 2 (o) and n ¼ 4 (x). KD or EC50 was 10 mM
24 3 Classical Analytical Solutions

Scheme (3.10) assumes simultaneous binding of n ligand molecules L to a

receptor R, without intermediates LR, L1R, L2R,. . . and the like. Such simultaneous
binding is unlikely in solution, but nevertheless (3.10) is commonly used for the
calculation of dose–response curves. It would lead to a binding function (3.11)

½Ln R] ¼ R0=ð1 þ ðKD /[L]Þn Þ (3.11)

KD is the equilibrium dissociation constant for each step, n is called the Hill
coefficient, and is usually denoted as nH. Unlike n in (3.10) the Hill coefficient n in
(3.11) need not be an integer. Mathematically (2.11) is a logistic function, which
originally had been developed [16] for the description of population growth.
Logistic functions are easy to calculate and easy to fit. They are often applied to
dose–response curves as “4 parameter logistic equations” (3.12)

Response ¼ (Max  Min)/(1 þ (EC50 /x)n Þ (3.12)

With x as the drug concentration and the four parameters, Max ¼ maximal
amplitude, Min ¼ background, EC50 ¼ ligand concentration for the half-maximal
response and n ¼ nH ¼ Hill coefficient, it has been pointed out that logistic
functions do not correspond to any realistic binding scheme unless the Hill coeffi-
cient is one [17, 18].
When logistic functions (3.12) are calculated for different Hill coefficients and
plotted versus the logarithm of the drug concentration, the result is a series of
symmetric sigmoid curves (Fig. 3.7). Such curves are very useful for fitting,
because the four parameters Max, Min, EC50 and nH are usually not correlated
(Sect. 8.1). Modification of nH changes the shape of the curve. Its maximal slope
thus can be varied independent of the inflection point (EC50). Amplitude (Max) and
background (Min) can also be varied independently.
From the logistic dose–response curves in Fig. 3.7, only the binding curve (+)
with a Hill coefficient of 1 corresponds to a plausible reaction scheme (3.1).
Experimental dose–response curves were found to be unsymmetrical when the
maximal slope was steeper than for the binding curve (+) shown in Fig. 3.7 [18].
For these types of experimental curves, there is no generally accepted explanation
[17, 18]. One new explanation for commonly observed steep dose–response curves
has been published [19]; another one is calculated from multiple allosteric
interactions (Fig. 5.12) and another one from irreversible reactions (Fig. 6.11).

3.2 Analytical Solutions for Binding Kinetics

The dissociation of ligand L from the receptor–ligand complex is a first-order

reaction as described in (2.6). Rearranging this equation gives

d[LR]/[LR] =  k1  dt (3.13)

3.2 Analytical Solutions for Binding Kinetics 25

This can be integrated from the initial concentration [LR]0 (the equilibrium
concentration for t ¼ 0 at the beginning of the dissociation reaction) to the concen-
tration of [LR] at time t:

½LR
ð ðt
d[LR=½LR ¼ k1  dt (3.14)
½LR0 0

The analytical solution is:

lnð[LR]Þ ¼ lnð[LR]0 Þ  k1  t (3.15)

[LR] ¼ [LR]0  ek1  t (3.16)

Equation (3.15) states that the logarithm of [LR] should be a linear function of t.
Dissociation can easily be measured when a large excess of a competing ligand is
added to the complex LR. Since the forward reaction is blocked by the competitor,
only the dissociation reaction with the rate constant k1 should be observed. This
type of experiment is extensively discussed and calculated in Sects. 6.4–6.6.

3.2.1 Exponential Decay

Equation (3.16) describes the exponential decrease of the complex [LR], and is
shown in Fig. 3.8. Equation (3.16) is based on (2.6) and therefore can be applied
whenever the decay rate of a substance is proportional to its concentration. Such a
phenomenon is rather common in nature, with radioactive decay as the most
prominent example.
The time which is required for the dissociation of half of the molecules is called
“half life” and often is denoted with the Greek letter t (tau). From (3.15) or (3.16)
one can compute that t is equal to ln2/k. It is independent of the initial concentra-
tion and therefore would be the same for the decay from 100% to 50% or from 50%
to 25% and so forth. A dissociation reaction followed for more than five half-lives is
shown in Fig. 3.7.
Measuring the half-life of a dissociation reaction is the simplest method for
obtaining the rate constant of this reaction: k ¼ ln2/t. It is also the simplest
method to ensure that a reaction truly is of the first order: Subsequent half-lives
must always give the same value, and one can begin to monitor a first-order reaction
at any time.
26 3 Classical Analytical Solutions

3.2.2 Half-Logarithmic Plots

Half-logarithmic plots of simple exponential decays should be linear, as they follow

(3.15). Experimental data often do not yield a linear half-logarithmic plot. In some
cases the dissociation reaction is more complex. In other cases, fluctuation of the
background signal may cause a deviation. Typically, one does not monitor absolute
concentrations, but will follow the reaction by a signal (fluorescence, NMR,
absorption,. . .) which is proportional to the concentration. Such a signal always
will have a background, and in order to take the logarithm of the signal in Fig. 3.8,
one has to take the logarithm of signal minus background. The background shown
in Fig. 3.8 would be the signal at “infinite” time, which translates to the signal after
at least ten half-lives. Fluctuations of the experimental points will lead to statistical
fluctuations, but fluctuations of the background will lead to systematic deviations as
shown in Fig. 3.9.
Drifts in the background can also be caused by typical experimental artifacts,
such as bleaching, instability of a lamp, temperature increase and so forth. The half-
logarithmic plot of ln [LR] vs. time (3.15) should be linear, but background
variations may lead to distortions (Fig. 3.9). This may be one of the reasons why
half-logarithmic plots have disappeared with the introduction of personal
computers. Instead, one typically would fit (3.16) and add the background signal
as an additional parameter. It should be stated, however, that the initial part of a

Exponential Decay (ana7.m)

100
Concentration of [LR] (bound ligand)

80

60

40

20

0
0 0.5 1 1.5 2 2.5 3 3.5 4
Time (seconds)

Fig. 3.8 Exponential Decay. Equation (3.16) is calculated with a rate constant k1 ¼ 1 s1, initial
concentration [LR]0 ¼ 100, t ¼ ln 2 s
3.2 Analytical Solutions for Binding Kinetics 27

Half-logarithmic plot (ana8.m)

4.5

3.5
ln (Concentration of [LR])

2.5

1.5

0.5

0
0 0.5 1 1.5 2 2.5 3 3.5 4
Time (seconds)

Fig. 3.9 Half-logarithmic plot of the reaction in Fig. 3.8. Equation (3.16) with k1 ¼ 1 s1,
[LR]0 ¼ 100 is calculated. () Logarithm of [LR]. (o) 2% background was added before the
logarithm was taken, (x) 2% background

complex reaction often is identical to one simple exponential curve and that
subsequent deviations may escape notice when background is used as a parameter
for curve fitting.

3.2.3 Initial Velocity of Association Kinetics

A reversible reaction may involve conformational changes and may be quite

complex, but the first part of an association reaction where a ligand L is added to
free receptor R (only one binding site) is simple at the time point t ¼ 0. When both
reaction partners meet for the first time, their initial free concentrations are equal to
the total concentrations:

v0 ¼ d[LR]/dtjt¼0 ¼ k1  L0  R0 (3.17)

Experimentally, the initial velocity is easy to measure, since it is the tangent to a

kinetic experiment at zero time. At zero time, there is no bound ligand formed so
that all back reactions can be ignored. When the concentration change of LR can be
quantified and when the initial concentrations L0 and R0 are known, the rate
constant k1 of the initial part of the association reaction can be determined without
curve fitting.
28 3 Classical Analytical Solutions

3.2.4 Pseudo First-Order Kinetics

The rate of a second-order reaction (2.1) is proportional to the concentration of both

reactants (2.2). If one of these concentrations is kept constant, the rate is only
proportional to the other. This results in a “pseudo” first-order reaction. If L0
is more than a factor of ten larger than R0, the free ligand concentration L0 of a
reversible reaction (2.9) does not change significantly upon binding and therefore can
be regarded as constant so that the differential equation for [LR] can be simplified:

d½LR/dt ¼ k1  L0  ½R  k1  ½LR

d½LR/dt ¼ k1  L0  ðR0  ½LRÞ  k1  ½LR (3.18)
d½LR/dt ¼ ðk1  L0 þ k1 Þ  ½LR þ k1  L0  R0

This is, apart from the constant k1 · L0 · R0, the same as (3.13). Integration
leads to

lnð[LR]Þ ¼ const  ðk1  L0 þ k1 Þ  t (3.19)

This is the same as an exponential decay (3.15) and (3.16) with an observed rate
constant k

k ¼ k1  L0 þ k1 (3.20)

If one measures binding kinetics at different concentrations of L0 (the ligand in

excess), one can obtain a series of pseudo first-order rate constants k. If one plots
these constants versus the concentrations of excess ligand, the result is a straight
line with a slope of k1 and an intersection with the y-axis of k1. It should be noted
that (3.20) is independent of the direction of the reaction. If the reaction is started
from a complex LR, and if this complex is diluted by a large factor, then the
observed dissociation also would follow first-order kinetics with the pseudo first-
order rate constant k from (3.19).

3.2.5 Multiple Exponential Fits

In many cases, binding kinetics is complex and cannot be fitted to simple

exponentials. Sometimes the following sum is used to fit the experimental data:

[LR] ¼ A1  ek1t + A2  ek2t + A3  ek3t þ . . . (3.21)

This is a sum of exponential functions with amplitudes A1, A2, A3,. . . and rate
constants k1, k2, k3,. . .. Almost all curves can be fitted with such a sum of
exponential curves. The resulting amplitudes and rate constants are not very
References 29

meaningful since generally there is no plausible reaction scheme which would lead
to (3.21). Sometimes (3.21) is used simply as a way to describe the data. This is
correct, as long as the whole ensemble of parameters (rate constants and
amplitudes) is reported. One should note that the parameters of (3.21) generally
are correlated so that individual rate constants cannot be extracted from a fit with
(3.21), unless they are significantly more than one order of magnitude apart.
Correlation is discussed in Sect. 8.1. Note that k1, k2, k3,. . . of a multi exponential
fit in a kinetic experiment are real numbers, and that (3.21) does not describe an
exponential Fourier series.
If an association reaction under pseudo first-order conditions does not give a single
exponential, then the reaction is more complex. The kinetics of such a model can only
be calculated from sets of differential equations and require numerical methods.

References

1. Langley JN (1905) On the reaction of cells and of nerve-endings to certain poisons, chiefly as
regards the reaction of striated muscle to nicotine and to curare. J Physiol 33:374–413
2. http://de.wikipedia.org/wiki/Scatchard-Diagramm
3. http://en.dogeno.us/2004/03/scatchard-plot/
4. Voet DJ, Voet JG (1995) Biochemistry. John Wiley & Sons, New York
5. Gutfreund H (1995) Kinetics for the life sciences. Receptors, transmitters and catalysts.
Cambridge University press, Cambridge
6. Segel LA, Slemrod M (1989) The quasi-steady-state assumption: a case study in perturbation.
SIAM Rev 31:446–477. doi:10.1137/1031091
7. Michaelis L, Menten ML (1913) Die Kinetik der Invertin-Wirkung. Biochem Z 49:333–369
8. Segel IH (1993) Enzyme kinetics: behavior and analysis of rapid equilibrium and steady-state
enzyme systems. Wiley Classical Library, New York
9. Copeland RA (2005) Evaluation of enzyme inhibitors in drug discovery: a guide for medicinal
chemists and pharmacologists. Wiley-VCH, Weinheim
10. Copeland RA (2000) Enzymes: a practical introduction to structure, mechanism, and data
analysis. Wiley, New York
11. Purich DL (2010) Enzyme kinetics: catalysis & control: a reference of theory and best-
practice methods. Elsevier, London
12. Cook PF, Cleland WW (2007) Enzyme kinetics and mechanism. Garland Science, New York
13. Leskovac V (2003) Comprehensive enzyme kinetics, Kindle Edition. Amazon
14. Lineweaver H, Burk D (1934) The determination of enzyme dissociation constants. J Am
Chem Soc 56:658–666
15. Hill AV (1910) The possible effects of the aggregation of the molecules of haemoglobin on its
dissociation curves. J Physiol (Lond) 40:iv–vii
16. Verhulst PF (1845) Recherches mathématiques sur la loi d’accroissement de la population.
Nouv mém de l’Academie Royale des Sci et Belles-Lettres de Bruxelles 18:1–41
17. Weiss JN (1997) The Hill equation revisited: uses and misuses. FASEB J 11:835–841
18. Prinz H (2010) Hill coefficients, dose-response curves and allosteric mechanisms. J Chem
Biol 3:37–44
19. Prinz H, Sch€onichen A (2008) Transient binding patches: a plausible concept for drug binding.
J Chem Biol 1:95–104
.
Chapter 4
Getting Started with Octave

GNU Octave is a free and open high-level language adequate for the calculation of
reaction schemes. Its installation is described in detail and the basic structure is
illustrated with the help of a very elementary tutorial. GNU Octave corresponds
strongly to MATLAB®. Both are matrix-oriented computer languages. The sample
programs often use matrices such as spreadsheets to calculate arrays of
concentrations. At the end of this chapter, the reader should have installed GNU
Octave and should be familiar with some basic features and the writing of simple
programs.

4.1 Installation Instructions

GNU Octave is available in the Internet. It is freely redistributable under the terms of
the GNU General Public License (GPL) [1] and ready to download at http://www.
gnu.org/software/octave/. For windows, use the windows installer from http://
octave.sourceforge.net/. This page from Octave Forge also provides assorted sets
of Octave functions, called “Packages”. We will need the package “Miscellaneous”,
the package “io”, and the package “Optim”. For windows version
3.2.3, the
packages are selected as “Choose Components” when setup of the windows installer
is executed. For other operating systems, the packages have to be downloaded sepa-
rately and linked with the Octave command pkg install. Detailed procedures for
Mac and Linux operating systems are described in the appendix. GNU Octave is
continuously being developed, so that the actual installation routines may change.
The same is true for MATLAB. This language is also developed continually.
Installation procedures may change, but a license has to be obtained before the
program is installed. Trial licenses may be available. Note that the MATLAB
Optimization Toolbox is required to solve nonlinear equations and to fit the data.
A statistical analysis of the fitted parameters with covariance and correlation
matrices requires functions such as nlinfit from the Statistics Toolbox. One
needs an additional license for each toolbox.

H. Prinz, Numerical Methods for the Life Scientist, 31

DOI 10.1007/978-3-642-20820-1_4, # Springer-Verlag Berlin Heidelberg 2011
32 4 Getting Started with Octave

Octave and MATLAB use very similar code. MATLAB or Octave or both often is
taught in university courses for mathematicians, physicists, or engineers. Both
languages are almost identical, so that someone who had learnt MATLAB will be
able to write code in GNU Octave, and vice versa. MathWorks Inc. offers excellent
seminars for MATLAB. The MATLAB programming environment is excellent, and
its debugging features are better than Octave’s. For larger programs, one may develop
the programs in MATLAB and distribute them to the scientific community in Octave.
For Octave, there are numerous tutorials and help pages available in the internet.
They are useful, but they should not be necessary for the understanding of the
sample programs presented in this book. These programs are introduced step by
step and all new commands are explained wherever they appear for the first time.
Notepad++ is a free windows editor available at http://notepad-plus-plus.org/. It
is useful for writing Octave code, but other editors may be equivalent or even better.
Useful editors should display line numbers, because error messages in Octave
usually refer to line numbers. Notepad++ can be configured to MATLAB/Octave
with the pull down tab Language/M/Matlab. Notepad++ then recognizes strings,
comments, instruction words, etc., and displays them in different styles. These
styles can be edited in Settings/Style Configurator/Matlab.

4.2 Typing the First Commands

GNU Octave is a modern high-level computer language, but looks “retro” for most
life scientists. There are (almost) no pull down menus, no buttons to click, and only
hidden help functions. The Spartan “terminal window” (Fig. 4.1), which appears
when Octave is executed in a Microsoft environment, may come as a shock, but one
can get used to it. There will be “cryptic” messages, a prompt (the “>” character),
and a blinking cursor. Nothing else!
Do not get intimidated. Just try it out and type
A ¼ 12*3
You will get the answer A ¼ 36 and again a prompt and a blinking cursor.
If you then type
A
Again you get the answer
A ¼ 36
However, if you type
a
you get an error message. Octave distinguishes between uppercase and lower-
case letters and does not “know” the variable a. If you type
B ¼ 2*4;
You get no answer. The semicolon is used to denote the end of a command and
prevents the direct output. If you then type A*B, (or c ¼ A*B), you will get
ans ¼ 288 (or c ¼ 288)
which is the product 36*8.
4.2 Typing the First Commands 33

Fig. 4.1 The Octave terminal window (Screenshot)

It should be emphasized that one cannot learn a computer language from a book
without applying the language. It is mandatory to run Octave at this stage and
one should “play” with it using different commands. Of course, all basic operations
(þ, , *, /,^) are the same as given on the numeric keyboard or on any
calculator. Play with it now!
The arrows of the keyboard help to navigate in the Octave terminal window:

Edit functions become available with a right mouse-click. This is the only pull
down menu available. Copy consists of “Mark” followed by highlighting the
characters to be copied, followed by “enter”, the return key. The standard windows
edit commands such as Ctrl + C or Ctrl + V do not work in the Octave terminal
window. Ctrl + C is reserved as a key to terminate any program. Therefore, one has
to use the right mouse-click for the edit functions.
Typing help followed by the name of a command will show help information
for the specified command. Usually, this help information is bigger than the
terminal window, and an information line is printed in the lower left corner of the
terminal window, giving the numbers of the displayed lines and the commands to
navigate: (f)orward, (b)ack, and (q)uit
Typing doc followed by an expression gives access to the GNU Octave Manual.
In this case, navigation is done with the keyboard. Ctrl + C will terminate the doc
session and return to the terminal window.
34 4 Getting Started with Octave

4.3 Writing a Program

Open your text editor and type these lines:

The line numbers are supplied by the editor, as shown in Fig. 4.2. They must not
be entered separately. Line numbers are useful for discussing and debugging the
program. If they are not shown in your editor, try the help function. All editors
should be able to display line numbers.
Do not forget that Octave is case sensitive. Once you have written the program,
you should give it a name like “start.m” and save it with that name. All programs
and functions in octave must have the ending “.m” This ending is derived
from mATLAB. Save the program into a working directory, such as C:/work or
C:/octave/work.
Then start the program GNU Octave. Once the terminal window appears, enter it
and type cd C:/work (or the name of your favorite working directory). The
command cd changes the directory to your folder. Once the prompt and the
blinking cursor appears, type the name of the program which you have just written
(without the ending .m), such as “start” in your terminal window. You will see a
new window and the plot shown in Fig. 4.3.
Now let us discuss the program start.m line by line. The first line assigns the
number 2 to the variable min. The equals sign (¼) in Octave and in any other
computer language is not symmetric. It takes the value of the right side and assigns
it to the variable on the left side. For example, if a ¼ 1 and b ¼ 2, then the

Fig. 4.2 The Notepad ++ editor window (Screenshot)

4.3 Writing a Program 35

Fig. 4.3 A plot generated by start.m. (Screenshot) The command plot(x,y) – in its most basic
variety – shows the data as dots connected with straight lines. The numbers in the lower right
corner correspond to the x and y coordinates of the cursor position (not visible in the screenshot)

command a ¼ b will result in a ¼ b ¼ 2, whereas b ¼ a will result in

a ¼ b ¼ 1.
Note that all lines end with a semicolon. If you omit the semicolon, the variable is
also sent to the terminal window. Try it out! The first three lines of start.m therefore
simply assign numbers to the variables min, max, and N. It is always a good idea to
define all variables and assign them to numbers in the beginning of the program, since
they are easy to find there. linspace in line 4 is an interesting function: It creates the
array x of N points, equally spaced linearly between min and max. You can see the
N ¼ seven components if you type “x” in the terminal window. In fact, you can see
the assigned value of any variable if you type in the variable name. Try it out with min
or N. Line 5 uses the operator.* (the character dot followed by an asterisk for the
multiplication sign). As explained in Sect. 4.6, this denotes element-by-element
multiplication of vectors (arrays are vectors in Octave). When the seven elements of
y (¼x2) are plotted versus the seven elements of x, using the plot command in line 6,
the result is a hyperbola. It will appear as “Figure 1” in a separate window.
The Octave plot function internally uses gnuplot, another software freely
distributable under the GPL. Information is available at www.gnuplot.info. The
sample programs shown below will increase in their complexity and will use more
and more features of gnuplot. Titles, axes, legends, and texts will be used, so that all
36 4 Getting Started with Octave

information will be stuffed into the plot outputs. Independent help pages, textbooks,
and tutorials are available for gnuplot. Just as for Octave, external help is useful, but
should not be required for the understanding of the sample programs here.

4.4 Setting Up a Work Environment for Octave: Arrangement

of Editor, Terminal, and Plot Window

Unlike MATLAB, Octave does not come with a cohesive work environment
containing all windows of interest. It is a good idea to arrange the relevant windows,
so that one can edit, run, and see the results of a program in one glance. Figure 4.4
shows such an arrangement.
Once the program start.m had been executed by typing start in the
terminal window, it can be run again with ", the up arrow followed by the return
key. Try it out. If the program start.m is modified (replace, for example, the
number 2 in line 1 with 10), it has to be saved again before it can be executed. Note
that the icon for the command “save” in Notepad++ turns red when a file has been
modified, but not saved.

Fig. 4.4 Arrangement of terminal, editor, and plot windows (Screenshot)

4.5 The MATLAB Environment 37

It is possible to copy single commands, or lines, or sections of a program from

the text editor to the terminal window, with Ctrl + C, but remember the unique
environment of the terminal window (right mouse-click and select “Paste” for the
paste command).

4.5 The MATLAB Environment

The MATLAB environment (Fig. 4.5) looks slightly different. There are pull down
menus, there are good help functions, and there are good debugging features. Still,
the setup is practically the same along with the language. The “Terminal Window”
from GNU Octave is called “Command Window” in MATLAB. It can be found
in the lower right corner, and it uses standard edit commands (Ctrl + C, etc.).
In MATLAB, all windows can be customized. A separate “Workspace” window
(upper left window) is useful, because the values of all variables are listed there.
There are many other user-friendly features of MATLAB, such as a separate help
window, but they are not covered here. MATLAB comes with extensive documen-
tation, but the basic features are almost [2] the same: One still has to write a
program in MATLAB/Octave code, and one has to write it line by line.

Fig. 4.5 One arrangement of MATLAB windows (Screenshot)

38 4 Getting Started with Octave

4.6 Arrays, Vectors, and Matrices

The name MATLAB is derived from MATrix and LABoratory. It is a matrix-

oriented language, and Octave has the same properties. Any variable initially is
regarded as a matrix, and a matrix with one column is a column vector, a matrix
with one row is a row vector, and a matrix with one element is a scalar. This may
sound complicated, but it really is simple once you understand the concept.
Matrices are written in square brackets.
Arrays of concentrations, such as the substrate concentrations of an experiment,
may be regarded as vectors. If one uses seven concentrations like 1, 2, 4, 10, 20, 40,
and 100 mM for an experiment, these concentrations would be entered as a state-
ment such as C0 ¼ [1, 2, 4, 10, 20, 40, 100] in Octave. Square brackets
define a matrix (in our case, a matrix with one row, which is a row vector). When
the elements are separated with a comma, it is a row vector, and when the elements
are separated with a semicolon, it is a column vector. The third concentration of the
array C0 is denoted as C0(3)in Octave. If you type this in the terminal window,
you will get the answer ans ¼ 4.
Let us consider two vectors, a row vector a ¼ [1,2,3] and a column vector
b ¼ [4;5;6] Octave is well suited for vector algebra, but the sample programs
listed in this book will use vectors mostly as arrays of concentrations. For these,
element-by-element multiplication is a useful operation, written as .* (dot
followed by an asterisk). The operation requires that the vectors contain the same
numbers of rows and columns. If you typed a.*b with the examples above, you
would get the error message “nonconformant arguments”, since a is a row vector,
whereas b is a column vector. The transpose operator is denoted as the character’
(the single quote or prime symbol). It transposes a row vector into a column vector
and vice versa. Therefore, the product a.*b’ will give the row vector [a1*b1,
a2*b2, a3*b3] for the above example. Try it out.
Note that the usual matrix multiplication is denoted by the operator “*” (aster-
isk). In the above example, a*b is the matrix multiplication of the row vector
a times the column vector b. The resulting scalar product is the number
a1*b1+a2*b2+a3*b3, which in our example is 32.
A matrix may be written as a column vector of row vectors or a row vector of
column vectors. M ¼ [a; a; a], for example, would be a 3X3 matrix where all
three rows contain the same elements a1, a2, and a3. The elements of a matrix can
be addressed directly as M(Nrow,Ncolumn) with Nrow and Ncolumn denoting
the index number for the row and the column, respectively. Vectors can be regarded
as matrices with one row or one column, respectively. For the row vector a, the
element a2 may be addressed as a2 ¼ a(2) ¼ a(1,2), and for the column
vector b, the third element might be addressed as b3 ¼ b(3) ¼ b(3,1).
At this stage, it is best to try out your own examples directly in the octave
window. If you have run the program start.m before, you may want to type y’
and see the result. Or type M ¼ [x; 2*x; 3*x] Or type 5*3 and get ans ¼ 15.
ans is the variable name (short for “answer”) for the result of the last octave
4.7 A Very Elementary Tutorial for GNU Octave and MATLAB 39

calculation where you did not supply a variable name. If you then type ans*3
afterward, you will get ans ¼ 45. Just try it out and play with it!

4.7 A Very Elementary Tutorial for GNU Octave

and MATLAB

It is easier to learn languages from practical exercises than from books alone. The
same is true for computer languages. One needs practical exercises. While we do
not want to limit the creativity of the reader, the commands listed below can be used
as a very elementary tutorial. If the commands (set in courier font) on the left are
typed directly line by line, either in the Octave terminal window or in the MATLAB
command window, the results will appear immediately.
40 4 Getting Started with Octave

The result of the typing should be a plot giving a parabola, a text within the plot,
Hallo as a title, and a label “World” for the x-axis. If all these commands are
understood, this tutorial has been finished successfully.
Some commands which may also be useful:

4.8 Recommended Literature for Octave/MATLAB

This will conclude a first elementary introduction to octave. There is extensive

literature and helpful information to be found abundantly in the Internet. Both GNU
Octave and MATLAB are living languages, which present their current
developments in the Internet, so that written textbooks may seem to be obsolete.
There are noteworthy exceptions: Cleve Moler has written an introductory course
on numerical methods [3], which gives the necessary background for numerical
methods applied here. Individual chapters can be downloaded from Mathworks.
com [4]. MATLAB comes with extensive documentation and help functions. For
GNU octave, its manual [5] is a useful piece of hardware. It gives an overview and
can be used as an excellent reference for each command. It is a useful investment,
although its contents are shown directly in the Octave terminal window with the
command doc. There is a very active GNU Octave community, so that there are
more than 326,000 Google hits for “Octave Tutorial”. Some of these seem to be
particularly useful [6–8], and most tutorials cover MATLAB as well. MATLAB is
at least as popular as Octave, with 571,000 Google hits for “MATLAB Tutorial”.
Tutorials may be available in any spoken language worldwide.
Octave and MATLAB have a broad application spectrum, mainly focused on
science and engineering. Only a few of these features are required for the calcula-
tion of reaction schemes. They are introduced one by one as they come along to
solve practical problems.

References

1. http://www.gnu.org/software/octave/license.html
2. http://en.wikibooks.org/wiki/MATLAB_Programming/Differences_between_Octave_and_MA
TLAB
References 41

3. Moler C (2010) Numerical computing with MATLAB. Society for Industrial Mathematics,
Philadelphia, PA
4. http://www.mathworks.com/moler/chapters.html
5. Eaton JW, Bateman D, Hauberg S (2008) GNU octave manual version 3. Network Theory Ltd,
London
6. http://smilodon.berkeley.edu/octavetut.pdf
7. http://en.wikibooks.org/wiki/Octave_Programming_Tutorial
8. http://srl.informatik.uni-freiburg.de/downloadsdir/Octave-Matlab-Tutorial.pdf
.
Chapter 5
Equilibrium Binding

Binding equilibria of n compounds can be calculated from n equations of n

unknowns. These are solved numerically with the Octave/MATLAB subroutine
fsolve. Writing program code for this task is introduced step by step. The output
of all programs is given in a graphic format so that binding mechanisms of
increasing complexity can be visualized. Allosteric interactions of subunits and
multiple allosteric interactions at one target molecule are calculated as practical
examples. At the end of this chapter, the reader should be able to calculate
equilibrium binding for any reaction scheme involving any number of ligands,
inhibitors and binding sites.

5.1 Solving Nonlinear Equations for Equilibrium Binding

Sections 2.2 and 2.3 describe the sets of nonlinear equations derived from equilib-
rium-binding schemes. In Octave and MATLAB, these equations are solved with the
function fsolve, which in turn is based on the MINIPACK subroutine hybrid [1].
MATLAB users must obtain the “Optimization toolbox,” which includes fsolve.
The set of equations to be solved has to be written as an array of n equations with
n unknowns. In Octave/MATLAB, such an array is treated as a vector, and the set of
equations can be written as a vector function

FðxÞ ¼ 0 (5.1)

F and x are vectors of the same size. The function fsolve requires an initial
estimate for the unknowns, a vector xo. It then calculates the set of equations as F
(x0). The result will not be zero, but fsolve varies the unknowns x in several steps
until (5.1) is solved within the required mathematical precision.
This general strategy has to be illustrated with an example; reversible binding to
one site leads to (2.16) and (2.17) for equilibrium binding. These equations can be
re-written as elements of the function (5.1):

H. Prinz, Numerical Methods for the Life Scientist, 43

DOI 10.1007/978-3-642-20820-1_5, # Springer-Verlag Berlin Heidelberg 2011
44 5 Equilibrium Binding

F(1) ¼ x(1) þ x(1)  x(2)/KD1  R0 ¼ 0 (5.2)

F(2) ¼ x(2) þ x(1)  x(2)/KD1  L0 ¼ 0 (5.3)

Remember, x(i) denotes the ith element of the vector of unknowns x, with
x(1) ¼ [R] and x(2) ¼ [L]. Likewise, the initial estimates will also be defined as
one vector x0. When R0 and L0 are similar, the free concentrations are estimated to
be half of the total concentrations:

x0ð1Þ ¼ 0:5  R0 (5.4)

x0ð2Þ ¼ 0:5  L0 (5.5)

Note: When the ligand concentration is much larger than the receptor concen-
tration, equation (5.6) becomes a better estimate for the free ligand concentration:

L0 >> R0 : x0ð2Þ ¼ L0  0:5  R0 (5.6)

Having explained the basic concept of iterative techniques used in fsolve, this
function can now be applied in a sample program. The function fsolve is called with
the statement: x ¼ fsolve('name',x0), where 'name' is the name of the
function for the set of (5.2) and (5.3). x0 and x are column vectors of the same
length. Remember, Octave and MATLAB are matrix-oriented languages, and all
variables may be matrices, vectors or scalars.

5.2 Equilibrium Binding to One Site (EQ1.m)

The first sample program is EQ1.m, a program which calculates reversible binding
to one site with analytical and numeric methods. The central function, called from
fsolve, is called EQ1F.m. It is stored with this name in the octave work directory.
It mainly contains (5.2) and (5.3) written in Octave code.

The first line is merely a description of the function. It is written as a “comment.”

Comments are defined as lines beginning with the percent sign %. In MATLAB, two
of these signs (line 1) define the beginning of a program section. The % sign may not
only appear at the beginning, as illustrated in lines 4 and 5. In this case, the
comment begins at the % sign and ends at the end of the line. The function
5.2 Equilibrium Binding to One Site (EQ1.m) 45

statement (line 2) defines the name of the function (EQ1F). The function names in
this textbook are derived from the names of the main program, supplemented by the
letter F. The name of a function must be the same as the name of the file in which it
is stored (EQ1F.m). The letter F in line 2 defines the variable name which is
returned from the function. The components of the vector F are calculated in lines 4
and 5. They correspond to (5.2) and (5.3). The global statement in line 3 allows
data exchange between the function and the main program.
The main program (EQ1.m) also begins with a short description given as
comment lines.

The next section of EQ1.m is the definition of all relevant parameters. It is

usually a good idea to define parameters at the beginning of a program. Computer
code is read from top to bottom so that any variable has to be defined before it is
used. When all variables are defined together in consecutive lines, it is easy to
identify them for debugging or modifying the program.

Clear in line 7 is a command which is useful at the beginning of any program,

since all user-defined variables are cleared from the memory. Octave remembers all
parameters, even after a program has successfully been terminated. For example, if
you run EQ1.m by typing EQ1 in the octave terminal window, you will see the plot
of Fig. 5.1 and the > character in the terminal window. When you then type R0 in
the terminal window, you will get R0 ¼ 1 as an answer. Liberating memory with
the clear command helps to keep the slate clear.
The string variable filename defined in line 8 will be used later in the program,
in lines 30, 36, 37, 38 and 40. A string variable is defined by the apostrophes in
line 8. The statement global in line 9 corresponds to the same statement in line 3
of the function EQ1F and allows transfer of the specified parameters. Lines 10–13
simply assign numbers to variable names. linspace in line 14 returns the row
vector VL0 with N linearly spaced elements between 0 and MaxLo. These 30 ligand
concentrations will be used for plotting the data. Variable names beginning with V
emphasize that the variable is a vector.
46 5 Equilibrium Binding

Equilibrium Binding curve (EQ1.m)

1
numeric
analytic
Concentration of [LR] (bound ligand)

0.8

0.6

0.4

0.2

0
0 20 40 60 80 100
Concentration of [L] (free ligand)

Fig. 5.1 Equilibrium-binding curve. Bound ligand versus free ligand is shown for 1 mM receptor
and an equilibrium dissociation constant of 10 mM. The values were computed with the program
EQ1.m as numeric approximations (x) and analytical solutions (+)

The main program basically contains a loop running from line 16 to line 25.
Loops are sets of commands which are executed over and over, until a condition is
fulfilled. The simplest condition is the number of repeats, which is N in line 16. The
command for i ¼ 1:N states that the index i begins at 1 and is increased by its
default value of 1 until i ¼ N. For each i, a different total ligand concentration L0
(line 17) is used from VL0, the vector of initial concentrations which had been
created in line 14. Lines 18 and 19 give the initial estimates and correspond to
(5.4) and (5.5). Line 20 calls the function fsolve. It requires the name of the
5.2 Equilibrium Binding to One Site (EQ1.m) 47

subroutine function 'EQ1F' where the equations are written. Note that the
name'EQ1F' is a string of characters as defined by single-quote ('). fsolve
also requires the vector x0 of initial estimates as an argument. The result in line 20
is the vector x. In lines 21 and 22, the elements of the vector x are translated into
the chemical notation as free receptor R and free ligand L, respectively. Line 23
defines VLR as the vector of complexes LR, employing (2.15). VL (line 24) simply
gives the vector of free ligand concentration L. The loop ends with an end
statement. Octave allows different end statements like endfor, endif,
endwhile and endfunction. Since MATLAB does not support this, we do
not make use of this Octave feature.
The loop creates the vectors VLR and VL of complex and free ligand, respec-
tively. Calculating the analytical solution in octave does not require a loop. Line 26
is octave code and looks identical to (3.1), but note that VL is a vector of
concentrations L. In line 26 all concentrations of bound ligand LR for all free
ligand concentrations L are calculated with one statement. The operator ./ (dot
followed by forward slash) in line 26 denotes an element-by-element division.
It corresponds to the element-by-element multiplication explained in Sect. 4.7.
Once all relevant concentrations are calculated, they can be plotted with the plot
command given in line 28. The first argument in the plot function is the x-value, in
our case the free concentration vector VL and the second is the corresponding y-
value, the bound ligand vector VLR. The style (included in quotes) specifies line ()
and symbol (x). A second pair of x and y coordinates is supplied for the analytical
solution (LR_analytical). This is specified by another symbol (+). Lines 29 and 30
are remarkable for two reasons. First, the continuation marker (three dots) at the end
of a line indicates that the next line belongs to the same statement. In this case, lines
29 and 30 are pasted to a single title function. The second remarkable feature of
the title function is the string which is given as an argument. Every string may be
regarded as a row vector of characters. In Octave, any row vector is defined by
its elements, separated by commas, in square brackets (Sect. 4.7). Therefore
['Equilibrium Binding curve (',filename,'.m)'] is a row vector of
the string 'Equilibrium Binding curve (', followed by the parameter string
filename (defined in line 8), followed by the string '.m)'.

The same procedure (row vector of strings) is used in the print and save
commands in lines 36, 37, 38 and 40. Unfortunately, Octave version 3.2.4 gives
a message implicit conversion from matrix to string whenever a
print command (line 36 and 37) is executed. Such a warning is not given when the
48 5 Equilibrium Binding

same program is run in MATLAB or older versions of Octave. Therefore, please

ignore the message implicit conversion from matrix to string
displayed in the Octave Terminal window of version 3.2.4.
The statement axis (line 31) requires a row vector of min and max values for
the x- and y-axes, respectively. The legend function (line 34) inserts a legend into
the plot, thus also relates to the plot function. The two strings correspond to the
two data pairs in the plot command of line 28. A third argument is a number (2, in
this case), which defines the quadrant in which the legend will appear.

The last lines of the program EQ1.m give examples for different output formats.
The print command in lines 36 and 37 is used to save the graphic output in two
graphic formats, namely .jpg (line 36) and .emf (line 37). The later format
(Microsoft-enhanced metafile) was used for all figures in this textbook. All
parameters can be saved with the save command in line 38. The data might be
retrieved at any time with a corresponding load command. Octave (not
MATLAB) stores the resulting data file EQ1_all.txt in an ASCII format
which can be accessed with any text editor. In both computer languages, the data
can be stored in ASCII format as a spreadsheet. This is done with the option
–ascii in line 40, but one can save only a single matrix with this command.
Therefore, the matrix out was generated in line 39 as a row of the column vectors
of interest. Figure 5.1 is generated when EQ1 is typed in the Octave terminal or
MATLAB command window. As mentioned before, please ignore the two
messages implicit conversion from matrix to string. They appear
when Octave version 3.2.4 is used.
How to modify the sample program. All programs in this book are examples,
which can and should be modified so that they may be used as templates for other
problems. One possible modification of EQ1.m is shown in the program EQ2.m,
which calculated equilibrium binding to two sites. But there are numerous other
possibilities. Variables can be changed in lines 10–14. The plot may be modified
in line 28, and title, legends and labels in lines 29–34. Try it out! For example, if
you want to show your data as a double reciprocal plot, you simply modify line 28.
Note, however, that one cannot simply use a division for vectors. One has to use
element-by-element division (./):

If you only changed line 28 and run EQ1 in the octave terminal window, you
will get an empty plot, because you forgot to modify the axis. One may remove line
31 and the plot function will find its own default axes. Instead of removing a line, it
5.2 Equilibrium Binding to One Site (EQ1.m) 49

is easier to define it as a comment by adding the % (percent) character at the

beginning of line 31 to define it as a comment:

And . . . voila, a double reciprocal plot appears when you enter EQ1 in the
terminal window. Of course, the axis labels are still wrong, but play with it. One
can easily modify everything.
The program EQ1b.m is a modification of EQ1.m for the calculation of the
difference between numerical and analytical solutions. As discussed in Sect. 2.6,
numerical solutions are approximations, whereas analytical solutions always are
correct. For Octave 3.2.3 and its default parameters the differences between numer-
ical and analytical solutions of EQ1.m are shown in Fig. 5.2.
The differences were computed with EQ1b.m, a program which differs from
EQ1.m mainly in lines 27 and 29.

Difference between numerical and analytical solutions (EQ1b.m)

8e-007

7e-007

6e-007

5e-007

4e-007

3e-007

2e-007

1e-007

-1e-007

-2e-007
0 20 40 60 80 100
Concentration of [L] (free ligand)

Fig. 5.2 Differences between numerical and analytical solutions. Bound and free ligand
concentrations shown in Fig. 5.1 were calculated for 1 unit (mM) receptor and an equilibrium
dissociation constant of 10 units (mM). The differences of the numerical and analytical solution as
calculated with Octave version 3.2.3 are plotted versus the free ligand concentration. Note the
scale of the y-axis, which merely covers a total range of 106 units, corresponding to 1 pm, which
is not visible clearly in Fig. 5.1
50 5 Equilibrium Binding

The differences between analytical and numerical solutions can, of course, be

neglected for all practical reasons. They are more than a million-fold lower than
the calculated values for bound ligand. If the accuracy of numerical solutions is
an issue, it can be optimized in MATLAB with the optimset command.
Figure 5.2 is shown here as a caveat. There may be a situation where numerical
solutions give unexpected results. We all have a tendency to believe in
computers, but numerical artifacts must not be ignored. If they are suspected,
they can be detected by switching between Octave and MATLAB or with
modifying the optimization parameters or even running the same program on a
different computer platform.

5.3 Equilibrium Binding to Two Sites (EQ2.m)

Reaction scheme (2.19) shows ligand binding to a receptor where two binding sites
are accessible. The corresponding set of equations is given in (2.24) and (2.25).
Obviously, they differ from the set of equations calculated in EQ1.m so that the
function EQ1F has to be modified. A modified program should have a new name,
and EQ2F is appropriate. Equations (2.24) and (2.25) translate to:

Please note a programming technique, which is unnecessary for the computer,

but makes life much easier for the programmer. When the equations become more
complex, it is a good idea to replace x(1) by R and x(2) by L in lines 6 and 7. Of
course, R and L have to be assigned, and this is done in lines 4 and 5 before they are
used for the equations.
The main program is stored as EQ2.m. Again the concentrations of complexes
are calculated inside a loop for each ligand concentration in lines 23–29, and the
plot command in line 32 can use the vectors generated within the loop.
5.3 Equilibrium Binding to Two Sites (EQ2.m) 51

The plot command is used to plot the concentrations of all complexes. This is
done in line 32. The legend in line 38 (not shown here) is changed accordingly.
Figure 5.3 shows the concentrations of LR, RL and LRL as a function of the free
ligand concentration. Note that the concentrations of the monoliganded complexes
LR and RL only differ by the same factor at all ligand concentrations. This is a
general property of all equilibrium-binding curves, where complexes with the same
number of ligands only differ by a given factor at all ligand concentrations. The
complex LRL initially shows a sigmoid increase. This has nothing to do with
cooperativity. It simply results from concentration dependence of the plot, where
monoliganded complexes LR and RL are favored at lower ligand concentrations
than the saturated complex LRL.
How to modify the sample program. The parameters in the program EQ2.m are
specified in lines 9–12, and can easily be modified. For extremely low receptor
concentrations, one may consider (5.6) and replace x0(2) ¼ 0.5*L0; in line 29
with x0(2) ¼ L0-0.5*R0;
One practical modification is shown in EQ2b.m: Most binding studies do not
measure the concentration of complexes, as in Fig. 5.3, but the concentration of
bound ligand. The concentration CB of bound ligand is

CB ¼ RL þ LR þ 2  LRL (5.7)

To plot this, the following lines are modified:

52 5 Equilibrium Binding

Equilibrium Binding - Two accessible sites (EQ2.m)

1
LR
RL
LRL
0.8
Concentration of Complexes

0.6

0.4

0.2

0
0 20 40 60 80 100
Concentration of [L] (free ligand)

Fig. 5.3 Equilibrium binding to two independent sites. For reaction scheme (2.19), the
concentrations of complexes LR (x) RL (+) and LRL (*) are plotted versus the free ligand
concentration. The calculated affinities are KD1 ¼ KD3 ¼ 5 mM for the first and KD2 ¼ KD4
¼ 10 mM for the second site

Lines 2 and 8 are required when the name of the program is changed. It is
advisable to rename a program every time it is modified. Line 31 is a translation of
(5.7) into Octave code. Line 33 is the modified plot command. Note that it is
possible to do simple operations like 2*VLR within the plot command, and also
note that it is possible to define colors in the plot style (defined with quotes). The
three symbols of '-ok' at the end of line 40 translate into –: solid line o: point
style as circle and k: color black. The colors are specified in Octave as 'k' black,
'r' red, 'g' green, 'b' blue, 'm' magenta, 'c' cyan and 'w' white. Figure 5.4
is a diagram of bound ligand rather than a diagram of the different complex
concentrations. Of course, the bound ligand of the complex LRL is 2·LRL.

5.4 Equilibrium Binding to Two Sites in the Presence

of Inhibitor (EQ3.m)

When a second ligand can also bind to the receptor, the reaction scheme looks
complicated, although the underlying mechanism is not. In most cases, the second
ligand is an inhibitor so that the second ligand is denoted as I in scheme (5.8). Just
5.4 Equilibrium Binding to Two Sites in the Presence of Inhibitor (EQ3.m) 53

Equilibrium Binding - Two accessible sites (EQ2b.m)

2
LR
RL
2*LRL
Bound
1.5
Bound Ligand

0.5

0
0 20 40 60 80 100
Concentration of [L] (free ligand)

Fig. 5.4 Ligand binding to a receptor with two sites. For reaction scheme (2.19), the concentration
of bound ligand (0) is plotted versus the free ligand concentration. The contribution of the
complexes LR (x) RL (+) and LRL (*) to the total binding curve is shown. The affinities in
(1.19) are KD1 ¼ 50 mM, KD2 ¼ 20 mM, KD3 ¼ mM and KD4 ¼ 12.5 mM

Equilibrium Binding - Two sites with Inhibitor (EQ3.m)

2
Concentration of bound ligand

1.5

0.5

0
0 20 40 60 80 100
Concentration of [L] (free ligand)

Fig. 5.5 Equilibrium binding – two sites with inhibitor. The concentration of bound ligand was
calculated from reaction scheme (5.8) with KD1 ¼ 10 mM, KD2 ¼ 5 mM, KD3 ¼ 5 mM, KI1
¼ 3 mM, KI2 ¼ 5 mM, KI3 ¼ 3 mM, KI5 ¼ 10 mM and KI6 ¼ 10 mM. The total inhibitor
concentrations are (from top to bottom) 0, 25, 50, 75 and 100 mM. The total receptor concentration
is 1 mM
54 5 Equilibrium Binding

like L in scheme (2.19), the inhibitor I may bind to two sites. In reaction scheme
(5.8), the assignment of the sites to the respective affinities is consistent with
scheme (2.19) so that the equilibrium dissociation constant KI1 for the inhibitor
concerns the “left” site, just like the equilibrium dissociation constant KD1 for the
ligand. There are two mixed complexes, LRI and IRL. LRI can either be formed by
the addition of L to RI (equilibrium dissociation constant KD6) or by the addition of
I to LR, with the equilibrium dissociation constant KI5.

ð5:8Þ

To simplify the reaction scheme, free ligands are not shown explicitly in (5.8).
Moreover, the reactions are not assigned with their rate constants, but simply with
numbers. The letter I in front of the reaction number indicates that inhibitor is added
at this reaction step.
There are four coupled equilibria, i.e. four closed circles in (5.8). Similar to
(2.22) the affinities of the three additional ternary complexes must be independent
of the reaction pathway so that in addition to (2.23), (5.9)–(5.11) must hold. This
reduces the number of independent variables from 12 to 8 independent equilibrium
dissociation constants.

KD 1  KD 3 ¼ KD 2  KD 4 (2.23)

KI 1  KI 3 ¼ KI 2  KI 4 (5.9)

KD 2  KI 6 ¼ KI 1  KD 5 (5.10)

KD 1  KI 5 ¼ KI 2  KD 6 (5.11)

The dependent equilibrium dissociation constants are calculated in lines 18–21

of the main program EQ3.m.
5.4 Equilibrium Binding to Two Sites in the Presence of Inhibitor (EQ3.m) 55

These four equilibrium dissociation constants are not required in EQ3F.m and
therefore need not be included in the global declaration in line 2 of EQ3F.m and
line 12 of EQ3.m. The main difference to EQ2.m is the additional ligand I which
translates to the additional independent variable x(3) in line 5 and to the addi-
tional equation F(3) in lines 11–12 of EQ3F.m:

The three equations with the three unknowns R, L and I in lines 6–12 follow the
pattern described in (2.31). For each of these equations, all (free, bound and total)
concentrations involving the respective unknown (R, L or I) of scheme (5.8) have to
be included. Each of the complexes required for the bound concentration is calcu-
lated by the law of mass action from free receptor R, free ligand L and free inhibitor
I concentration. Note that bound ligand calculated from LRL in line 9 involves the
factor 2, since two ligand molecules are bound in that complex. The same is true for
IRI in line 11.
The binding curves are calculated with arrays of 30 ligand and 5 inhibitor
concentrations as defined in lines 26–31. These concentrations are varied in a
nested loop with the index i varying between 1 and 30 and the index k varying
between 1 and 5. The loops run between lines 33 and 54.
56 5 Equilibrium Binding

The concentrations of all eight complexes as well as the free ligand and inhibitor
concentrations are calculated in lines 44–53 within the loop. The bound ligand
BoundL can then be calculated outside the loop. Note that the concentrations
generated inside the loop are not just vectors (arrays) of concentrations at different
total ligand concentration, but now are matrixes (spreadsheets) of concentrations at
different total ligand and total inhibitor concentration. The elements of these
matrices are identified with the indices i and k. For spreadsheets, rows are typically
addressed with numbers and columns with letters. A matrix element VRL(3,2)
therefore would correspond to the cell B3 in a spreadsheet. For readers familiar with
spreadsheets, this relation may help to identify the matrix elements.
It has been mentioned in Chap. 4 that every variable in Octave is treated by default
as a matrix. The statement in line 56 therefore is matrix addition, so each element of
the spreadsheet VRL is added to each corresponding element of VLR and so forth.
This leads to rather elegant programming code. Likewise, the plot command in line
56 looks elegant and simple, but it plots five columns of bound ligand calculated for
five different inhibitor concentrations versus the corresponding columns of substrate
concentration. The results are shown in Fig. 5.5.
Increasing the number of ligands and/or the number of sites and/or the number of
conformational states can be done in two steps. First, one has to make a list of all the
expected complexes and calculate them as products of their free concentrations
divided by the respective equilibrium dissociation constants, just like those in lines
6–8 of EQ3F.m. With an increasing number of complexes it may be more
complicated to draw a reaction scheme than to write down these equations. The
second step consists of writing one equation of the type (2.31) for each ligand. The
resulting set of equations F(x) has to be solved by fsolve in the main program.
5.5 Allosteric Interactions of Subunits (EQ4.m) 57

This procedure can get as complex as one likes, but always follows the same
pattern.
The main advantage of numerical methods and a high-level programming
language consists in its flexibility. One can calculate the concentration dependence
of any complex or any set of complexes. For example, fluorescence resonance
energy transfer (FRET) is specific for the interaction of ligand and inhibitor and can
be employed to study its mechanism [2]. Part of the signal often is proportional to
the concentrations of ternary complexes, where both types of ligands are bound to
the receptor simultaneously. Modifying the program EQ3.m and plotting the
ternary complexes versus the substrate concentration is performed with the program
EQ3b.m and the statements below:

In line 56 the mixed ternary complexes IRL and LRI are added, because they
show the same dependence on inhibitor and ligand concentration. The function max
() in lines 57 and 61 is worth mentioning. For a matrix, it returns the maximal
values for each column in a matrix. Therefore, the result of line 56 is a row vector
MA with the maximal values of mix calculated at the five different inhibitor
concentrations. For a vector argument, the function max in line 61 simply returns
the maximal value so that max(MA) is the maximal value of all ternary complexes
at all inhibitor concentrations. The resulting plot is shown in Fig. 5.6.
The concentrations of the ternary complexes IRL and LRI alike depend on both
the ligand and the inhibitor concentrations. At high ligand concentration, the ligand
competes with the inhibitor and forms LRL, and at high inhibitor concentrations,
the inhibitor mainly binds as IRI. Therefore, the concentration dependence of the
ternary complexes is not easy to predict and has to be calculated. Note the circles on
the x-axis. They should correspond to the concentration of ternary complexes in the
absence of inhibitor and should all be zero. With the numerical solution returned
from the Octave algorithm, some were slightly negative and thus could not be
plotted within the axes defined in line 61. This problem had been addressed in
Fig. 5.2 and discussed in Sect. 2.6.

5.5 Allosteric Interactions of Subunits (EQ4.m)

The term “allosteric” is used differently in the literature [3]. For most cases today, it
states that a ligand may bind to a protein and elicit a conformational change. This
change may then change the properties of the protein at another (“allosteric”) site.
In contrast, the concerted model for oxygen binding to hemoglobin of Monod,
Wyman and Changeux [4] (“MWC” model) concerns the allosteric interactions of
58 5 Equilibrium Binding

Equilibrium Binding - Two sites with Inhibitor (EQ3b.m)

0.2
Concentration of bound ligand

0.15

0.1

0.05

0
0 20 40 60 80 100
Concentration of [L] (free ligand)

Fig. 5.6 Concentration of ternary complexes IRL + LRI calculated from reaction scheme (5.8)
with KD1 ¼ 10 mM, KD2 ¼ 5 mM, KD3 ¼ 5 mM, KI1 ¼ 3 mM, KI2 ¼ 5 mM, KI3 ¼ 3 mM,
KI5 ¼ 10 mM and KI6 ¼ 10 mM. The total inhibitor concentrations are (from top to bottom) 25,
50, 75 and 100 mM. The total receptor concentration is 1 mM

protein subunits. It assumes that the four hemoglobin subunits may exist in two
conformational states (R and T) even in the absence of ligand as shown in reaction
scheme (5.12).

ð5:12Þ

These conformational states traditionally are named R (for relaxed) and T (for
tense). In the absence of ligand, the subunits are predominantly in the R conforma-
tion ([R] > [T] translates to KD2 > 1 in scheme (5.12)). If the ligand L has a higher
affinity for T (KD3 < KD1) then the equilibrium of these conformational states is
shifted upon binding. The binding equilibrium of reaction scheme (5.12) is calcu-
lated from the set of equations given in EQ4a.m:
5.5 Allosteric Interactions of Subunits (EQ4.m) 59

Line 6 calculates the equilibrium concentration of T so that this conformational

state can then be used in lines 7 and 8. These equations again calculate the sums of
all (free, bound and total) receptor and all (free, bound and total) ligand
concentrations. The program EQ4a.m solves equations 7 and 8 and calculates a
binding curve as a Scatchard plot (Fig. 5.7)
Figure 5.7 illustrates a general principle of equilibrium binding; no matter how
many conformations are there, they will always be in equilibrium at any given ligand
concentrations. The concentrations of conformational states differ by the same factor
at all ligand concentrations in equilibrium studies. Consequently, since only one
binding site is involved, the Scatchard plots are linear for the bound ligand ([LT] +
[LR]), as well as for the complexes [LR] and [LT]. Reaction scheme (5.12) shows
coupled equilibria with two conformational states, but only one site. The cooperative

Scatchard Plot - Two conformations R and T (EQ4a.m)

0.012
Bound
LR
LT
0.01
Concentration bound / free ligand

0.008

0.006

0.004

0.002

0
0 0.02 0.04 0.06 0.08 0.1
Concentration of bound ligand

Fig. 5.7 Scatchard Plot: Equilibrium binding to a receptor in two conformational states, R and T.
Reaction scheme (5.12) was calculated with KD1 ¼ 100 mM, KD2 ¼ 100 mM, KD3 ¼ 100 nM
and a receptor concentration of R0 ¼ 100 nM. The ligand L0 concentration was varied from 0.1 to
100 nM. Total bound ligand (o) as well as the two single bound complexes LR (x) and LT(+) are
shown in the Scatchard diagram
60 5 Equilibrium Binding

binding, which is the main point of the MWC model [4], is not contained within
scheme (5.12). For this, an allosteric interaction of subunits is required.
For simplicity, we will assume that the subunits R and T can form dimers (not
tetramers, as calculated with the MWC model for hemoglobin). These dimers may
be formed independently of the occupation of the monomers with ligand so that
16 dimers have to be considered. Four of these (RR, RT, TR and TT) are formed in
the absence of ligand. Depending on the symmetry of the molecule, the dimer TR
may be different from the dimer RT so that both dimers have to be distinguished.

ð5:13Þ

The allosteric monomer–dimer interaction is usually regarded to be independent of

the occupation of the monomers so that four equilibrium dissociation constants
KD5–KD8 are sufficient to calculate dimer formation in (5.13). For cooperative
binding, one would expect KD6 and KD7 to be large so that the dimers of the type
RT and TR are unlikely. The elements R, T, LR and LT of the dimer formation in
reaction scheme (5.13) depend on the ligand concentration L as shown in (5.12). The
reaction schemes (5.12) and (5.13) therefore have to be merged for calculating all
equilibrium complexes. The function EQ4F.m becomes rather complex, but the
principle behind is simple. For example (5.13) states that the concentration of LTRL is

½LTRL ¼ ½LT  [LR]/KD 7 (5.14)

from (5.12)

[LR] ¼ [L]  [R]/KD 1 (5.15)

[LT] ¼ ½L  [T]/KD 3 (5.16)

Therefore:

[LTRL] ¼ [L]  [L]  ½R  [T]/(KD 1  KD 3  KD 7Þ (5.17)

All other complexes are calculated with the same principle, applying the law of
mass action. Function EQ4F looks complex, because all 16 complexes of (5.13)
have to be considered.
5.5 Allosteric Interactions of Subunits (EQ4.m) 61

The factor 2 is required when there are two receptor subunits or two ligand
molecules contained within a complex. In the main program (EQ4.m), the bound
ligand is calculated from the sum of all these complexes:

The resulting Scatchard plot (Fig. 5.8) not only shows the total bound ligand
from lines 41 to 42, it also includes the ligand bound in the complexes LT and
LTTL, as stated in lines 44 and 45.
62 5 Equilibrium Binding

Scatchard Plot - Allosteric MWC Model for 2 sununits (EQ4.m)

0.007
Bound
LT
0.006 2*LTTL
Concentration bound / free ligand

0.005

0.004

0.003

0.002

0.001

0
0 0.02 0.04 0.06 0.08 0.1
Concentration of bound ligand

Fig. 5.8 Scatchard Plot: Allosteric Model. Reaction scheme (5.13) with (5.12) was combined in
EQ4F.m and calculated with KD1 ¼ 100 mM, KD2 ¼ 100 mM, KD3 ¼ 100 nM and a receptor
concentration of R0 ¼ 100 nM. RR, RT, TR and TT interactions were calculated with equilibrium
dissociation constants of 100 nM, 100 mM, 100 mM and 100 nM, respectively. The ligand
concentration L0 was varied from 0.1 to 100 mM. Total bound ligand (o), [LT] (x) and
2·[LTTL] (+) are shown in the Scatchard diagram

Figure 5.8 shows a curved Scatchard plot not only for the totally bound ligand,
but also for the main components. This clearly shows cooperative binding which is
induced by the allosteric interaction of subunits (5.13). Comparing Fig. 5.8 to
Fig. 5.7 illustrates that cooperative binding can only be observed when more than
one binding site is involved. In the case of the MWC model [4], the number of sites
corresponds to the number of interacting subunits, not to the number of ligand-
binding sites on one subunit.
How to modify program EQ4.m. The parameters can be changed in lines 10–17.
As soon as the equilibrium dissociation constants of dimer formation KD5–KD8 are
equal, the cooperativity is lost. It may be interesting to calculate the ratios of R/T, of
RL to TL and of LRRL to LTTL as a function of ligand concentrations. It certainly
will help to understand the intrinsic properties of the MWC model.

5.6 Allosteric Activators and Inhibitors (EQ5.m)

As mentioned before, there are many other interpretations of the word “allosteric,”
mainly referring to allosteric interactions between ligands and proteins, not to the
allosteric interactions between subunits [4] as described above. A dedicated review
5.6 Allosteric Activators and Inhibitors (EQ5.m) 63

[6] explains the different usage of the term. Here we do not discuss molecular
interpretations, but calculate reaction schemes instead. The simplest feasible
scheme (5.18) may be derived from the following reasoning; the word “allosteric”
is used when one wants to state that another (an “allosteric”) ligand may influence
the binding of the main ligand without direct competition. In this case, the binding
of the allosteric effector would influence the conformation of the receptor molecule
and either enhance the activity of the primary ligand or decrease it. In the first case,
the effector would be called an “activator,” and in the second case it is an inhibitor.
Interaction between subunits is not required. In the simplest form, a corresponding
reaction scheme is depicted in (5.18), where L is the ligand, R the receptor and I the
allosteric effector.

ð5:18Þ

Scheme (5.18) corresponds to equilibrium binding in the presence of inhibitor.

If, however, KD4 < < KD2, the “inhibitor” (the effector I), will stabilize LRI. I
then increases the affinity of L to RI (coupled equilibrium (5.11)) so that I is an
allosteric activator and no inhibitor. The set of equations for this scheme is given in
EQ5F.m:

When the bound ligand is calculated in EQ5.m as the sum of LR and LRI in
reaction scheme (5.14), the Scatchard plot (Fig. 5.9) can be calculated. It is linear
for all effector concentrations, indicating that the effector changes the affinity, but
that it cannot lead to cooperativity. Figure 5.9 was calculated with KD1 ¼ 10 mM
and KD3 ¼ 1 mM so that the allosteric effector I acts as an activator. For KD1 <
KD3, it acts as an inhibitor, because the ligand binding to RI has a lower affinity
than the ligand binding to the free receptor R.
64 5 Equilibrium Binding

Scatchard Plot: Allosteric Effector at one site (EQ5.m)

1
1 µM
10 µM
100 µM
Concentration of (bound / free) ligand

1 mM
0.8

0.6

0.4

0.2

0
0 0.2 0.4 0.6 0.8 1
Concentration bound ligand

Fig. 5.9 Scatchard plot in the presence of different concentrations of an allosteric effector.
Reaction scheme (5.18) was calculated with KD1 ¼ 10 mM, KD2 ¼ 100 mM, KD3 ¼ 1 mM
and a receptor concentration of 1 mM. The effector concentration is varied from 1 (bottom line)
to 1,000 mM (top line), indicating an increase in apparent affinity with the effector concentration

Figure 5.9 shows a linear Scatchard plot for an allosteric activator. A linear
Scatchard plot would also be observed for KD1 < KD3, i.e. for an allosteric
inhibitor. This concludes that cooperative binding is not related to allosteric
mechanism per se. Cooperative effects can only be expected when a ligand or an
inhibitor or an activator binds to more than one site of a receptor molecule. For the
MWC model, the additional sites are generated by allosteric interactions between
subunits, leading to multiple sites on the holoreceptor.

5.7 Dose–Response Curves (EQ6.m)

Dose–response curves are basically equilibrium-binding curves, plotted in a loga-

rithmic scale (Sect. 3.1.5). The response (be it enzymatic activity, the opening of an
ion channel, the cellular activity or any physiological response) is regarded as the
result of agonist binding to a receptor. In a logarithmic scale, any equilibrium-
binding curve and any dose–response curve looks sigmoid (Figs. 3.7 and 5.10).
5.7 Dose–Response Curves (EQ6.m) 65

Dose-response curves: Allosteric Inhibitor (EQ6.m)

1
1 µM
10 µM
100 µM
1 mM
0.8
Bound Ligand (µM)

0.6

0.4

0.2

0
0 0.5 1 1.5 2 2.5 3 3.5 4
log (concentration of inhibitor (µM))

Fig. 5.10 Dose–response curves (bound ligand vs. log inhibitor concentration) for different
agonist concentrations. Reaction scheme (5.18) was calculated with KD1 ¼ 10 mM, KD2 ¼ 10
mM, KD3 ¼ 1 mM and a receptor concentration of 1 mM. The ligand concentration is varied from
1 (bottom line) to 1,000 mM (top line) in four steps

For screening campaigns, the response resulting from agonist binding simply is
used as a signal, and the inhibition or stimulation of this signal as a function of drug
concentration is investigated. Most compounds in screening campaigns act as
inhibitors so that the activity typically is reduced with increasing drug (inhibitor)
concentration. This type of dose–response curve is shown in Fig. 5.10, calculated
for different ligand concentrations of reaction scheme (5.18). The equations for this
scheme are given in the function EQ5F.m. The parameters used for Fig. 5.10 imply
inhibition, whereby the allosteric inhibitor decreases the affinity for the ligand, but
does not block it completely. In terms of reaction scheme this corresponds to a
factor 100 between the affinity of the agonist (L) in to R in the presence (KD3) or
absence (KD1) of inhibitor.
The extrapolations of the response to zero inhibitor concentrations shown in
Fig. 5.10 reflect the formation of LR at the different ligand concentrations, whereas
the extrapolations to infinite inhibitor concentrations reflect the formation of LRI.
In screening campaigns, one usually does not plot bound ligand in absolute
scales, but shows the data as percent of maximal–minimal signal. This type of
signal manipulation is performed in lines 42–55 of the program EQ6b.m.
66 5 Equilibrium Binding

For zero inhibitor concentrations, the maximal concentration VLRmax (line 43)
is calculated from the function EQ1F.m with the equilibrium dissociation constant
KD1. For infinite inhibitor concentrations, VLRmin (line 47) is calculated from the
same function (EQ1F.m, binding to one site in the absence of an inhibitor) with the
equilibrium dissociation constant KD3. EQ1F expects KD1 as the parameter name
to be transferred via the global statement so that KD3 has to be renamed KD1 in
line 45. The variable temp is used in line 44 to save the value of KD1 and restore it
in line 48. With VLRmax and VLRmin, the response can then be calibrated as
percent values in lines 53 and 54.
With these modifications, the dose–response curves from Fig. 5.10 all have the
same shape, as shown in Fig. 5.11. They are shifted along the x-axis in agreement
with competition mechanisms, where the apparent KD for an inhibitor is increased
with the ligand concentration. The shape of these dose- or inhibitor–response
curves corresponds exactly to simple binding curves, which look sigmoid in a
logarithmic scale, as shown in Fig. 3.7. The Hill coefficient of all these curves is
one. Again, the allosteric model of (5.18) cannot explain the results of screening
campaigns, where dose–response curves often show steeper maximal slopes,
corresponding to Hill coefficients larger than one [7].
5.8 Multiple Allosteric Inhibition (EQ7.m) 67

Normalized dose-response curves: Allosteric Inhibitor (EQ6b.m)

100
1 µM
10 µM
100 µM
80 1 mM

60
% response

40

20

0
0 0.5 1 1.5 2 2.5 3 3.5 4
log (concentration of inhibitor)

Fig. 5.11 Normalized dose–response curves (% activity vs. inhibitor concentration) for different
ligand concentrations. Reaction scheme (5.18) was calculated with KD1 ¼ 10 mM, KD2 ¼ 10
mM, KD3 ¼ 1 mM and a receptor concentration of 1 mM. The ligand concentration is varied from
1 (bottom line) to 1,000 mM (top line)

5.8 Multiple Allosteric Inhibition (EQ7.m)

Dose–response curves are measured under equilibrium condition, or at least are

intended to be measured under equilibrium conditions (but note Fig. 5.11). When
any concentration of a complex is measured under equilibrium conditions as a
function of other concentrations, its shape is only influenced by the number of
binding sites, not by the number of conformations. Therefore, inhibitor–response
curves under equilibrium conditions with a steeper maximal slope than shown in
Fig. 5.11 can only be envisioned when more than one inhibitor-binding site is
involved in the process. Reaction scheme (5.19) depicts the simplest of these cases.
A ligand may bind specifically to one site (the active site) of a receptor, but there
may also be different sites for inhibitors. If the binding of one or more of inhibitors
leads to loss of the active conformation of the receptor, then the ligand may lose its
affinity to the active site. In the case of enzymes, loss of the active conformation
corresponds to loss of activity.
Denaturation via conformational changes is rather trivial. Heating of proteins
leads to conformational changes and typically to inactivation. Denaturation by the
binding of molecules is also not so very new, since protons, ionic detergents, salts or
68 5 Equilibrium Binding

organic solvents, etc. all bind to sites different from the ligand-binding site and lead
to denaturation and inactivation of proteins. These are unspecific allosteric
mechanisms. Drugs identified in screening campaigns are specific for their target
proteins, yet they may act via conformational changes at multiple sites. Reaction
scheme (5.19) shows an example for a maximum of six multiple allosteric sites.

ð5:19Þ

Reaction scheme (5.19) need not imply a mechanism with exactly six sites. It
can be used to calculate inhibitor binding to one site when only KI1 is near the
inhibitor concentration, and the other equilibrium dissociation constants for the
inhibitor are orders of magnitude larger. Likewise, it can be used to calculate
inhibition with 2, 3, 4, 5 and a maximum of six sites, depending on the chosen
constants.

The set of equations from (5.19) are calculated with the function EQ7F.m, listed
below. The binding of inhibitors in scheme (5.19) is a simplified sequential
mechanism. The sequential equilibrium dissociation constants KI correspond to
the minimum of parameters for the respective number of sites and therefore are
useful for the calculations. With the assumption of accessible and equivalent sites,
intrinsic equilibrium dissociation constants can be calculated with the help of
(2.30).

The main program, EQ7.m, initially defines KI1 ¼ 100 mM and assumes that
all other equilibrium dissociation constants for the inhibitor (KI ) take the extremely
5.8 Multiple Allosteric Inhibition (EQ7.m) 69

high value of 1 M (lines 16–21). This combination of parameters allows the

calculation of inhibition with one inhibitor site. This is done in the loop in lines
25–26. In line 37 the command KI2 ¼ KI1 sets KI2 ¼ 100 mM so that the next
loop in lines 37–49 can calculate inhibition with two inhibitor sites and so forth.
The program gets rather long because six of these loops have to be executed. This
type of programming does not look elegant, but copy and paste is fast and effective.

Figure 5.12 shows inhibitor–response curves for 1, 2, 3 and 6 sites with the same
equilibrium dissociation constant of 100 mM. Of course, the shapes of these curves
not only depend on the number of inhibitor sites, but also on their relative affinities.

Multiple allosteric inhibition (EQ7.m)

100
1:1 inhibition
2:1 inhibition
3:1 inhibition
80 6:1 inhibition

60
% response

40

20

0
0 0.5 1 1.5 2 2.5 3 3.5 4
log (concentration of inhibitor (µM))

Fig. 5.12 Multiple allosteric inhibition. Dose–response curves for relative activity vs. inhibitor
concentration were calculated for 1, 2, 3 and 6 multiple allosteric inhibitor sites. Receptor
concentration R0 ¼ 1 mM, ligand (agonist) concentration L0 ¼ 10 mM and KD1 ¼ 10 mM. All
equilibrium dissociation constants from scheme (5.19) were calculated as 100 mM with (+) 1, (x) 2,
(o) 3 and (*) six binding sites for the inhibitor
70 5 Equilibrium Binding

Different equilibrium dissociation constants for the inhibitors can be set in lines
16–21 of EQ7.m.
The calculated curves shown in Fig. 5.12 are not symmetric. They differ
markedly from symmetric logistic functions (Fig. 3.7) routinely used for the
analysis of dose–response curves [8]. The asymmetric shape is a general feature
of compound binding to multiple sites: At low concentrations, only a small fraction
of the sites is occupied, mostly in the form of singly bound receptor. For
dose–activity curves this results in relatively low increases of inhibition with the
inhibitor concentration. Multiple bound receptors will only appear at higher inhibi-
tor concentrations, where the concentration of free (active) receptor then rapidly
decreases.
Reaction scheme (5.19) or dose–response curves shown in Fig. 5.12 do not prove
allosteric interactions. The same scheme and the same dose–response curves have
been interpreted isosterically as transient binding patches [9]. We can conclude that
dose–response curves with Hill coefficients larger than one cannot be calculated
from binding equilibria with one inhibitor or activator binding site alone, be it an
allosteric mechanism or not. For our own studies with chitinase inhibition, we had
found Hill coefficients different from one, looked again at our X-ray structures and
identified two inhibitor-binding sites [10]. Whenever Hill coefficients larger than
one are found, there must be reasons for it. Fitting dose–response curves to
plausible reaction schemes should be possible with the programs presented here
or with enhanced modifications. Additional evidence (like isothermal calorimetry
or crystal structure) will help to refine the model and lead to a consistent under-
standing of the mechanisms involved.

References

1. More JJ, Garbow BS, Hillstrom KE (1980) User guide for MINIPACK-1. Argonne National
Laboratory Report ANL-80-74
2. Berger W, Prinz H, Striessnig J, Kang H-C, Haugland R, Glossmann H (1994) Complex
molecular mechanism for dihydropyridine binding to L-type Ca2+-channels as revealed by
fluorescence resonance energy transfer. Biochemistry 33:11875–11883
3. http://en.wikipedia.org/wiki/Allosteric_regulation
4. Monod J, Wyman J, Changeux JP (1965) On the nature of allosteric transitions: a plausible
model. J Mol Biol 12:88–118
5. Verhulst PF (1845) Recherches mathématiques sur la loi d’accroissement de la population.
Nouv mém de l’Academie Royale des Sci et Belles-Lettres de Bruxelles 18:1–41
6. Fenton AW (2008) Allostery: an illustrated definition for the ‘second secret of life’. Trends
Biochem Sci 33:420–425
7. NIH Data base: http://www.ncbi.nlm.nih.gov/sites/entrez. Chem BioAssay
8. Guidelines for standardized dose-response curves: http://www.ncgc.nih.gov/guidance/
manual_toc.html
9. Prinz H, Sch€onichen A (2008) Transient binding patches: a plausible concept for drug binding.
J Chem Biol 1:95–104
10. Pantoom S, Vetter IR, Prinz H, Suginta W (2011) Potent family-18 chitinase inhibitors: X-ray
structures, affinities and binding mechanisms. J Biol Chem, 286:24312–24323
Chapter 6
Binding Kinetics

Differential equations are solved by calculating the difference quotient and adding
the computed concentration differences to the initial values. This is a typical task
for a computer. For GNU Octave, the function lsode is used as a universal solver for
ordinary differential equations, while a selection of different solvers is available in
MATLAB®. They are all used with a similar syntax. The main challenge for the
scientist consists in transforming reaction schemes into differential equations. From
these, association or dissociation kinetics is computed by selecting different initial
concentrations. Sample programs for lag-phase, facilitated dissociation, sequential
binding mechanisms or irreversible inhibitors are explained. At the end of this
chapter, the reader should be able to write and solve differential equations for any
reaction scheme, be it as complex as desired.

6.1 Solving Differential Equations in GNU Octave

and MATLAB

Numeric methods are ideally suited for the calculation of differential equations.
Mathematically, the differential quotient dx/dt is the limiting value of the difference
quotient Dx/Dt for infinitesimal small differences. Computer programs simply use
their power of repetitive commands: Calculate the differences Dx of all concentrations
for a small time interval Dt, add these differences to the original values and do the
calculations for the next time interval, until the desired time range is covered.
This procedure is called a numerical solution of differential equations. For
ordinary differential equations (“ODE”s, they can be written as dx/dt ¼ f(x, t);
all feasible reaction schemes fall into this category) Octave and MATLAB provide
procedures to solve them. For Octave, the solver is the function lsode. It is based
on ODEPAC [1] a collection of Fortran solvers. In MATLAB, there is a family
of similar solvers called ode23, ode45, ode113, ode15s, ode23s,
ode23t, ode23tb.

H. Prinz, Numerical Methods for the Life Scientist, 71

DOI 10.1007/978-3-642-20820-1_6, # Springer-Verlag Berlin Heidelberg 2011
72 6 Binding Kinetics

All of these solvers require initial concentrations, because all differences Dx

have to be calculated from starting values. Such initial concentrations are no
estimates, and therefore should not be confused with x0 in fsolve. In Octave and
MATLAB, the solvers require that the differential equations are supplied as
functions. Again (like in fsolve in Chap. 5), the set of equations (6.1) to be
solved is written as a vector, whereby each element of this vector is a differential
equation.

Dx ¼ fðxÞ  Dt (6.1)

with Dx, f and x as vectors. Since the solvers have their own algorithms to optimize
the values for Dt, the functions for Octave and MATLAB solvers are written
without Dt as

Dx ¼ f(xÞ (6.2)

Both, for Octave and for MATLAB solvers, these vectors are column vectors.
In GNU Octave, the solver is called with the command

ð6:3Þ

In MATLAB, the corresponding command is

ð6:4Þ

ode45 is a versatile function from the MATLAB library of solvers. For both
computer languages, t is a vector of time points for which the differential equations
have to be solved, and x0 is a column vector of all initial concentrations. M is the
matrix of the results. Its rows correspond to the time points, and the columns to the
different complexes. Therefore, calculating concentration changes of 3 molecules
at 30 time points will give a matrix of 30 rows and 3 columns. T in (6.4) is basically
the same vector as t, but whereas t may be a column or a row vector, the resulting T
in MATLAB always is a column vector. The main (and trivial) difference between
GNU Octave and MATLAB, however, is the order of arguments in the functions
[(6.3) (6.4)]. This corresponds to reversed orders of arguments in the function
name, when called either from lsode in Octave or from ode45 in MATLAB.
In GNU Octave, the basic syntax is:

ð6:5Þ

In MATLAB, the corresponding function has to have the arguments exchanged:

ð6:6Þ
6.2 Kinetics of Ligand Binding to One Site (kin1.m) 73

This is all, but unfortunately it implies that Octave functions cannot be called
from ode45 in MATLAB, and that MATLAB functions cannot be called from
lsode in GNU Octave. Therefore, all programs involving differential equations
and all corresponding functions have to be supplied in two versions, one for each
programming language. The MATLAB versions are marked as nameM.m and
stored in the MATLAB directory, whereas the Octave versions are not distin-
guished and listed as name.m, stored in the Octave directory.

6.2 Kinetics of Ligand Binding to One Site (kin1.m)

These general considerations have to be substantiated with a simple example.

Reversible binding of the ligand L to a receptor R [reaction scheme (2.9)] leads
to the formation of the complex LR and to a corresponding decrease of L and R.
The difference quotients are:

D[L]/Dt ¼ k1  [L]  [R] þ k1  [LR] (6.7)

D[R]/Dt ¼ k1  [L]  [R] þ k1  [LR] (6.8)

D[LR]/Dt ¼ k1  [L]  [R]  k1  [LR] (6.9)

These equations correspond to lines 5–7 in the Octave function kin1F.m. The
command zeros in line 4 defines a column vector dx with 3 elements. The
numerical value (zero) is not important at this stage, and line 4 is only used to
define the dimensions of the output dx.

For MATLAB, x and t are exchanged in their position in line 2:

Everything else remains the same. Note that x(1) ¼ [L], x(2) ¼ [R] and
x(3) ¼ [LR]. The differential equations kin1F are solved with lsode in
line 26 of the main program kin1.m:
74 6 Binding Kinetics

The time points t are defined in line 23, the initial concentrations x1 as a column
vector in line 24. The initial concentrations for all reactants are denoted as L1, R1
and LR1. We may avoid the more common L0, R0 and LR0 at this stage, because
the zero has been used for total concentrations in Chap. 5. Note that the time points
given in the row vector t (lines 23, 26 and 31), only are those time points, for
which the calculated values are shown, and not all the infinitesimal small time steps
which are computed internally.
The solution of the differential equations is given as a matrix M, which is a row
of column vectors for all the computed concentrations. These concentrations are
assigned to their symbols in lines 27–29. The colon (:) in line 27 translates to the
command: “Take all rows (the colon character stands for the whole range) from
column 1 of the matrix M and assign it to the new column vector L”. The length of
these new column vectors L, R and LR is the same as the number of time points t.
The results are shown in Fig. 6.1.
Running the program kin1.m in MATLAB only requires that line 26 is
changed. The MATLAB version of kin1.m is named kin1M.m, and line 26 of
the MATLAB version is the following:

The resulting matrix M in MATLAB has the same structure as in Octave, so that
no other lines have to be modified in the MATLAB version.
Figure 6.1 shows the association of ligand L (+) and receptor R (x) as a function
of time. Both of these free concentrations decrease, while the concentration of
bound ligand LR (*) increases. Unlike pseudo-first order kinetics described in
Sect. 3.2.4, numerical methods allow this reaction to be computed at any ratio of
ligand to receptor, even when it is almost 1:1, as shown here.
How to modify the sample program. Just by changing the initial concentrations,
dissociation kinetics can be computed with the same program which had been used
for association kinetics. The program kin1b.m gives such an example, with the
initial concentrations defined in lines 18–20.
6.2 Kinetics of Ligand Binding to One Site (kin1.m) 75

Association kinetics of binding to one site (kin1.m)

1
[L]
[R]
[LR]
0.8
Concentration (µM)

0.6

0.4

0.2

0
0 20 40 60 80 100
time (sec)

Fig. 6.1 Association kinetics of binding to one site. Reaction scheme (2.9) is calculated with
k1 ¼ 0.02 mM1 s1 and k1 ¼ 0.001 s1. The initial ligand and receptor concentrations are 0.8
and 1 mM, respectively. Free ligand L (+), free receptor R (x) and the complex LR (*) are shown as
a function of time
76 6 Binding Kinetics

The resulting plot is shown in Fig. 6.2. It corresponds to a hypothetical dissocia-

tion experiment from a complex LR1 ¼ 1.0 mM, with free ligand concentration
L1 ¼ 0 and free receptor concentration R1 ¼ 0.1 mM. Kinetic experiments are
calculated from initial concentrations, and the initial concentrations of all reaction
partners have to be defined. This contrasts to the equilibrium binding studies of
Chap. 5, where all equilibrium concentrations of all complexes are computed
only from the total concentrations. The total concentrations in kin1b.m are
L0 ¼ L1 + LR1 ¼ 1.0 mM and R0 ¼ R1 + LR1 ¼ 1.1 mM.
Figure 6.2 shows the increase in free ligand (+) and free receptor (x)
concentrations together with the expected concentration decrease of the complex
(*). The relatively large amplitude results from the assumption that the free ligand
concentration is zero at the beginning of the reaction. This is an exercise to
demonstrate that the same differential equations (kin1F) are used to calculate
association (Fig. 6.1) and dissociation (Fig. 6.2) alike. It does not correspond to a
real dissociation experiment, where the initial concentrations would have to be
computed from an initial equilibrium.

Dissociation kinetics for binding to one site (kin1b.m)

1
[L]
[R]
[LR]
0.8
Concentration (µM)

0.6

0.4

0.2

0
0 20 40 60 80 100
time (sec)

Fig. 6.2 Dissociation kinetics of binding to one site. Reaction scheme (2.9) is calculated with
k1 ¼ 0.02 mM1 s1 and k1 ¼ 0.01 s1. The initial complex and receptor concentrations are
1 and 0.1 mM, respectively. The ligand concentration at zero time is zero
6.3 Binding to One Site Followed by a Conformational Change of the Receptor (kin2.m) 77

6.3 Binding to One Site Followed by a Conformational

Change of the Receptor (kin2.m)

In most cases, binding of a ligand will lead to conformational changes of the target
protein. Such conformational changes cannot be detected in equilibrium binding
studies, but often are detected in kinetic experiments. In the simplest case, they
follow reaction scheme (6.10):

ð6:10Þ

Scheme (6.10) can be translated into the four differential equations

(6.11)–(6.14). Four different components of the reactions (L, R, LR and LR*) are
involved, so that four differential concentration changes have to be computed.

d[L]/dt ¼ k1  [L]  [R] þ k1  [LR] (6.11)

d[R]/dt ¼ k1  [L]  [R] þ k1  [LR] (6.12)

d[LR]=dt ¼ k1  [L]  [R]  k1  [LR]  k2  [LR] þ k2  [LR  (6.13)

d[LR  ]/dt ¼ k2  [LR]  k2  [LR  (6.14)

These equations are solved with the octave program kin2.m. The function
kin2F.m contains the differential equations (6.11)–(6.14). The function has the
same name as the main program, completed with the letter F. The complex LR* is
named LRS (LR star) in octave code because special characters cannot be used in
variable names. The function kin2F.m uses L, R, LR and LRS in lines 9–12 for
the sake of clarity. These variable names have to be translated from the vector
components x(i) in lines 5–8. Likewise, the computed concentration changes
have to be translated into the vector notation dx(i) in lines 13–16. This ensures
together with line 4 that the output of the function kin2F.m is a column vector of
four elements.
78 6 Binding Kinetics

Association kinetics is calculated from kin2F.m when the initial free

concentrations R1 and L1 are set equal to the total concentrations R0 and R0.
Then, the main characteristics of the conformational change LR ! LR* become
obvious in Fig. 6.3. The final product [LR*] is formed only after the initial binding

Association kinetics and induced conformational change(kin2.m)

1
[L]
[R]
[LR]
[LR*]
0.8
Concentration (µM)

0.6

0.4

0.2

0
0 20 40 60 80 100
time (sec)

Fig. 6.3 Association kinetics and induced conformational change. Reaction scheme (6.10) is
calculated with k1 ¼ 0.05 mM1 s1, k1 ¼ 0.001 s1, k2 ¼ 0.02 s1 and k2 ¼ 0.002 s1. The
initial ligand and receptor concentrations are 0.8 and 1 mM, respectively
6.3 Binding to One Site Followed by a Conformational Change of the Receptor (kin2.m) 79

process. Its concentration as a function of time (o) in Fig. 6.3 follows a sigmoid
curve. This time phase is called the “lag-phase”, because the formation of LR* lags
behind the decrease of L and R. Transient complexes such as [LR] sometimes can
only be inferred from such a lag-phase of the final product.
Figure 6.3 shows that the complex LR is formed transiently. Its decrease
correlates with the increase of LR*, the final product. The decrease of L and R
resembles exponential decay, but analyzing this as a sum of exponentials would be
deceptive.
How to modify the sample program. Rather than choosing arbitrary concen-
trations for the calculation of a hypothetical dissociation experiment, we will now
calculate dissociation observed after dilution. The initial concentrations given in lines
20–23 and the rate constants in lines 16–19 therefore are not modified. Instead,
the association reaction of Fig. 6.3 is calculated until equilibrium is reached at
tmax ¼ 10,000 s (line 24). After that, dissociation is initiated by diluting the
solution by a factor 100. For this, all concentrations have to be divided by the same
factor, as shown in lines 30–33. These diluted concentrations are the initial
concentrations of the dissociation kinetics. The modified program is called kin2.m.
80 6 Binding Kinetics

The last time point t(N) of the association reaction is used for the calculation of
the initial concentrations of the dilution experiment, and the respective equilibrium
concentrations are L2 ¼ M(N,1), R2 ¼ M(N,2), LR2 ¼ M(N,3) and LRS2 ¼
M(N,4). Check it by typing M in the Octave terminal window. The vector x2
consists of these concentrations calculated at the end of the association reactions. It
is used for the initial concentrations of the dissociation kinetics in line 35.
Alternatively, one could calculate the binding equilibria with fsolve from a set
of nonlinear equations as shown in Chap. 5, and then take these values as the initial
concentrations of the dilution kinetics. Both methods are equivalent, provided that a
real equilibrium is reached. But one should note that for most experiments the
incubation times are known, even when it is not clear if the equilibrium has been
reached. Therefore, calculating kinetics for a defined long time (like in kin2b.m)
usually corresponds to the experimental set-up.
Figure 6.4 shows dissociation kinetics induced by dilution. It corresponds to a
shift from one equilibrium to another. The new equilibrium has 100-fold lower
concentrations for all components of the reaction, so that decrease of LR* (o) is
expected. Of course, the increase in free receptor (x) and ligand (+) concentration in
Fig. 6.3 is calculated after the initial 100-fold dilution step. The transient concen-
tration change observed for LR is not visible any more, since the equilibrium
between LR and LR* is exactly KD2, for the old and for the new equilibrium

Dilution from an induced conformation (kin2b.m)

0.008
[L]
[R]
0.007 [LR]
[LR*]
0.006
Concentration (µM)

0.005

0.004

0.003

0.002

0.001

0
0 2000 4000 6000 8000 10000
time (sec)

Fig. 6.4 Dilution from an induced conformation. Reaction scheme (6.10) is calculated with
k1 ¼ 0.05 mM1 s1, k1 ¼ 0.001 s1, k2 ¼ 0.02 s1 and k2 ¼ 0.002 s1. The initial ligand
and receptor concentrations had been 0.8 and 1 mM, respectively. The reactions initially were
calculated to proceed for 10,000 s. After that time, all concentrations were diluted by a factor of
100 and calculated from then on as shown
6.4 Dissociation Kinetics: Chase with Inhibitor (kin3.m) 81

alike. This can be verified by typing LRS./LR in the octave or in the MATLAB
command window after kin2b has run. Comparing Figs. 6.4 and 6.3 reveals that
the dissociation kinetics is not simply the reverse association kinetics.

6.4 Dissociation Kinetics: Chase with Inhibitor (kin3.m)

Dilution experiments are easy to perform, but the results often are noisy. Dilution
typically leads to a decrease of signal by the dilution factor, so that one may wish to
study the reaction with a high dilution factor. But high dilution results in low signal
and bad signal to noise ratio. Low dilution results in high signal but small signal
change. This dilemma can be avoided when a competing ligand is added in large
excess. For a reversible reaction (2.9) large excess of inhibitor removes all free
receptor, so that only a first order dissociation of the bound ligand LR with the rate
constant k1 should be observed. This type of “chase” experiments can be calcu-
lated from scheme (6.15)

ð6:15Þ

Written in Octave code, the differential equations for competitive binding to one
site (6.15) are part of the function kin3F.m:

For calculating a chase experiment, one has to calculate the initial equilibrium of
ligand and receptor (2.9) first. For the program kin3.m, this is done with the
function EQ1F.m described in Sect. 5.2. The dissociation kinetics is initiated by the
addition of inhibitor. The decrease of [LR] (bound ligand) is shown in Fig. 6.5 for
different inhibitor concentrations. Lines 15–45 of kin3.m are listed below. The
first part up to line 27 defines the parameters. Note line 15: It defines global
parameters for two functions: The function EQ1F.m, which is used to calculate the
binding equilibrium, only needs KD1, R0 and L0. For the differential equations
defined in kin3F.m the additional variables k1, km1, ki1 and kim1 are listed
82 6 Binding Kinetics

in the global statement of line 15. This illustrates that any variable which is listed
in the global statement of the main program can be used in the global
statement of any function, independent of the order in which it appears.

Lines 23 and 24 illustrate an important point which will be addressed in more

detail in Sect. 8.1: Differential equations for reaction schemes are calculated from
forward and backward rate constants. However, these rate constants usually are
correlated when used for data fitting. The most important parameter is their
quotient, either the equilibrium dissociation constant KD ¼ k/k+ or its reciprocal
value, the equilibrium constant. In life sciences, the equilibrium dissociation
constant is more common, since it immediately gives the concentration of half-
maximal saturation at equilibrium conditions. Therefore, the KD values and associ-
ation rate constants are taken as parameters, and the dissociation rate constants are
calculated from these in lines 22 and 23. Likewise, competing ligands may have
similar association rates to the ligands themselves. Equal association rate constants
6.4 Dissociation Kinetics: Chase with Inhibitor (kin3.m) 83

Chase with a competitive inhibitor (kin3.m)

1
I0 = 100 µM
I0 = 400 µM
I0 = 700 µM
0.9 I0 =1000µM
Bound ligand (µM)

0.8

0.7

0.6

0.5
0 20 40 60 80 100
time (sec)

Fig. 6.5 Chase with a competitive inhibitor. Reaction scheme (6.15) is calculated with KD1 ¼
10 mM, KI1 ¼ 100 mM, k1 ¼ ki1 ¼ 0.001 mM1 s1. The initial concentrations are L0 ¼ 100 mM,
R0 ¼ 1 mM. The inhibitor concentrations are 100 (+), 400 (x), 700 (*) and 1,000 mM (o)

are stated line 21. Changes in the numerical value for k1 (defined in line 19) will
then also affect ki1.
The initial equilibrium is calculated with fsolve in line 31 from total ligand
and receptor concentrations L0 and R0. From this the initial free and bound ligand
and receptor concentrations L1, R1, LR1 are derived in lines 32–34. The initial
concentration of bound inhibitor IR1 is zero (line 35). Four total inhibitor
concentrations are defined with the help of linspace(100, 1000, 4) in line
27. Each of these is used as an initial free (¼total) inhibitor concentration I1 in
line 37 within the for loop (lines 36–41). The subsequent concentration changes
are computed from the differential equations with the function lsode in line 39.
The concentration of the bound ligand [LR] as a function of time is given in the
third column of the result matrix M. For each k within the loop, a column vector for
[LR] is added to the matrix LR. The result is a matrix LR with four columns for each
total inhibitor concentration. Type LR in the octave window and you can see its
structure. Lines 42–45 (after the loop) compute four concentration arrays (vectors)
for the bound ligand [LR] at the four different inhibitor concentrations from the
matrix M.
Figure 6.5 shows a decrease of bound ligand with time after the addition of
inhibitor. Note that the final concentration of bound ligand decreases with increas-
ing inhibitor concentration. This is an expected feature of competitive inhibition.
84 6 Binding Kinetics

Chase with excess of competitive inhibitor (kin3b.m)

1
I0 = 100 mM
I0 = 400 mM
0.9 I0 = 700 mM
I0 =1000mM

0.8
Bound ligand (µM)

0.7

0.6

0.5

0.4

0.3
0 20 40 60 80 100
time (sec)

Fig. 6.6 Chase with excess of competitive inhibitor. Reaction scheme (6.15) is calculated with
KD1 ¼ 10 mM, KI1 ¼ 100 mM, k1 ¼ ki1 ¼ 0.001 mM1 s1. The initial concentrations are
L0 ¼ 100 mM, R0 ¼ 1 mM. The inhibitor concentrations are 100 (+), 400 (x), 700 (*) and
1,000 mM (o)

How to modify the sample program. Again, it may be interesting to do the

calculations with different inhibitor concentrations and/or different rate constants.
Increasing the inhibitor concentrations by three orders of magnitude gives the
dissociation kinetics of Fig. 6.6. The rate determining step at high inhibitor
concentrations is the first order dissociation from [LR] to [L] + [R].
Figure 6.6 shows the dissociation kinetics after addition of high excess (100 mM
to 1 M) of inhibitor. All these curves are identical. The final value of [LR] at infinite
times will approach zero, since all receptor will be bound to inhibitor and form the
complex IR. This type of reaction is expected, and the rate constant of the resulting
first order reaction is k1.

6.5 Facilitated Dissociation (kin4.m)

Inhibitors which only bind to the inhibitor binding site (6.15) are called competitive
inhibitors. In many cases, inhibitors may bind to entirely different or to overlapping
sites. Independent of the structural model, a ternary complex LRI of ligand,
receptor and inhibitor may be formed according to reaction scheme (6.16)
6.5 Facilitated Dissociation (kin4.m) 85

ð6:16Þ

Note that reaction scheme (6.16) is similar (5.18), but that the rate constants for
ligand and inhibitor are systematically differentiated. Reaction scheme (6.16)
shows a closed loop, so that the ternary complex LRI may be formed by different
pathways. Its energy must be the same, so one of the four equilibrium dissociation
constants can be calculated from the three others in (6.17)

KD 1  KI 2 ¼ KD 2  KI 1 (6.17)

The differential equations for scheme (6.16) can be written directly in Octave
code

When a chase experiment (dissociation after addition of inhibitor) is calculated

from these equations, the initial concentrations of all components of the reaction
must be known. The initial equilibrium is calculated from scheme (6.16) without
inhibitor, and this corresponds to simple reversible binding (2.9), so that the
function EQ1F.m can be used as in kin3.m.
For comparison, kin4.m uses the same rate constants and concentrations as
kin3.m. For the additional reactions involving the ternary complex LRI, we can
choose three parameters: The two forward rate constants k2 and ki2 and one
equilibrium dissociation constants KD2 or KI2 [note (6.17)]. If we assume that all
forward rate constants are the same, this leaves the affinity of the ternary complex
as the only significant parameter. For overlapping sites, this affinity should be
several orders of magnitude lower than the affinity of the corresponding binary
complex. The calculations performed in kin4.m assume that this difference is
three orders of magnitude.
Compared to competitive inhibition (Figs. 6.5 and 6.6), Fig. 6.7 shows a
dramatic increase in dissociation rates. It reflects the opening of a new dissociation
pathway for L from LR to LRI to IR and has been named “facilitated dissociation”
[2]. Note that not only the amplitudes, but also the curvature of the dissociation
curves increase with the inhibitor concentration. Facilitated dissociation as
86 6 Binding Kinetics

Chase for overlapping sites (kin4.m)

1
I0 = 100 µM
I0 = 400 µM
I0 = 700 µM
0.9
I0 =1000µM

0.8
Bound ligand (µM)

0.7

0.6

0.5

0.4
0 20 40 60 80 100
time (sec)

Fig. 6.7 Chase for overlapping sites. Reaction scheme (6.16) is calculated with KD1 ¼ 10 mM,
KI1 ¼ 100 mM, KI2 ¼ 100 mM, k1 ¼ ki1 ¼ k2 ¼ ki2 ¼ 0.001 mM1 s1. The initial
concentrations are L0 ¼ 100 mM, R0 ¼ 1 mM. The inhibitor concentrations are I0 ¼ 100 (+),
300 (x), 700 (*) and 1,000 (o) mM. The ternary complex LRI has a low affinity, which may be
explained by overlapping sites

calculated here is has been reported before [3, 4]. It can be expected when the
ternary complex has a very low affinity. This may either result from overlapping
sites [3] or from a specific molecular mechanism involving considerable conforma-
tional changes [4]. The limiting rate constant at extremely high inhibitor
concentrations is k2, for the dissociation of LRI to RI. Facilitated dissociation
may escape our notice, because it corresponds to “common sense”. We seem to
expect that a ligand dissociates faster the more competitors are added, but of course
first order dissociation on its own cannot be faster than [LR]·ek·t, the first order
dissociation of the ligand shown in Fig. 6.6. Facilitation of this reaction by an
inhibitor requires at least a ternary complex.
How to modify the sample program. It may be interesting to calculate the
dissociation over a wider concentration range of inhibitor or for different inhibitor
affinities, but most importantly for different affinities of the ternary complex. This
easily is done by modifying lines 16–27 of kin4.m. The results can be understood
with the help of reaction scheme (6.16): The formation of LRI from IR is a second
order reaction and its rate increases with the inhibitor concentration. The dissocia-
tion of LRI to IR is a first order reaction. Therefore, the rate-limiting step for
facilitated dissociation is k2, which is related to the affinity of the ternary complex
LRI.
6.6 Sequential Binding and Obstructed Dissociation (kin5.m) 87

6.6 Sequential Binding and Obstructed Dissociation (kin5.m)

When ligands bind to a receptor, one usually assumes that the sites are randomly
accessible like in reaction scheme (2.19). This need not be the case. For example,
when two sites are located inside a pit or a channel, the ligand which binds first
may not be able to dissociate before the second ligand has been released [reaction
scheme (2.18)]. Equilibrium measurements cannot distinguish between random
and sequential mechanisms, since only the numerical values of the apparent disso-
ciation constants would change in accordance with (2.30). Kinetic measurements
can, however, detect such a sequential mechanism when a chase experiment is
performed [5]. In the presence of inhibitor, the reaction scheme of a sequential
mechanism (2.18) has to be extended as shown in reaction scheme (6.18).

ð6:18Þ

Note that there is only one complex LR in reaction scheme (6.18) and that the
ternary complex is denoted as LLR in order to account for the order of the bound
ligand. An inhibitor which follows the same binding scheme may also form the
complexes IR and IIR. There are two markedly different ternary complexes
formed with ligand, inhibitor and receptor. From the complex ILR the ligand can
only dissociate once the inhibitor has been released, whereas the ligand has to
dissociate first from the complex LIR. Some of the differential equations were
getting rather long, so that four lines were cut with the help of the . . . (three dots)
continuation marker.
88 6 Binding Kinetics

Dissociation kinetics can only be calculated from these equations when the
correct initial concentrations are known. The initial concentrations L1, R1, LR1
and LLR1 are calculated in lines 43–47 from binding equilibria with fsolve and
the function kin5EQF.m. The initial concentrations of complexes with inhibitor
(lines 48–51) are, of course, zero before the inhibitor is added.

The dissociation kinetics was then calculated within a loop (lines 52–66) for
three different inhibitor concentrations. The total bound ligand is calculated in line
65, as a matrix CB of column vectors for the different inhibitor concentrations k.
The result can be seen immediately if one types plot(t,CB) in the Octave or
MATLAB command windows. The three vectors for the three concentrations are
displayed as a function of time. The more complex plot command in line 71 is used
to introduce the symbols, so that the three inhibitor concentrations can be distin-
guished when Fig. 6.8 is printed in black and white.
6.6 Sequential Binding and Obstructed Dissociation (kin5.m) 89

For MATLAB, the solver ode45 becomes very slow when applied to reaction
scheme (6.18). The reason lies in multiple fast reversible reactions which underlie
the slow obstruction mechanism. Such a problem is commonly called “stiff” and
MATLAB has a large selection of solvers which deal with stiff differential
equations. The names of these MATLAB solvers end in “s” (for stiff), such as
ode15s or ode23s. The MATLAB version kin5M.m of the program kin5.m
90 6 Binding Kinetics

Ordered binding and obstucted dissociation (kin5.m)

2
I0 = 1 mM
I0 = 10 mM
I0=100 mM

1.5
Bound ligand (µM)

0.5

0
0 20 40 60 80 100
time (sec)

Fig. 6.8 Ordered binding and obstructed dissociation. Reaction scheme (6.18) (sequential binding
of ligand and inhibitor) is calculated with KD1 ¼ KD2 ¼ KD3 ¼ KI1 ¼ KI2 ¼ KI3 ¼ 10 mM,
k1 ¼ ki1 ¼ k2 ¼ ki2 ¼ 1 mM1 s1. The initial concentrations are L0 ¼ 100 mM, R0 ¼ 1 mM.
The concentrations of the chasing ligand are 1,000 (+), 10,000 (x) and 100,000 (*) mM

therefore employs the solver ode23s instead of ode45, which had been used in
the previous programs. The Octave function lsode used in line 55 is quite capable
to handle these types of stiff differential equations.
The resulting dissociation curve (Fig. 6.8) is clearly biphasic. In the first part, the
ligand bound to the second site rapidly decreases. In molecular terms, the ligand L
dissociates from LLR to LR. All off-rates were assumed to be 10 s1, which is three
orders of magnitude faster than the off-rates assumed for the calculations in
Figs. 6.4–6.7. The fast off-rates were chosen because the second part of the
dissociation kinetics reveals a unique property of this mechanism: The dissociation
kinetics become slower the more inhibitor is added. The inhibitor binds to LR and
forms ILR. The sequential mechanism does not allow the dissociation of L from
ILR. I have to dissociate first, leading to LR, which again may react rapidly with I.
Thereby L is trapped. Such a sequential mechanism had been detected from this
type of experiment for the nicotinic acetylcholine receptor from electrophorus
electricus [5].
How to modify the sample program. The fast phase of the biphasic dissociation
in Fig. 6.8 is not resolved. A shorter time range is set with the command tmax ¼ 1;
in line 36. The axes are chosen by the plot routine, when the axis command in line
75 is defined as a comment with the help of %, the percent character.
Figure 6.9 shows that for high inhibitor concentrations the initial phase of the
dissociation reaches a maximum velocity. This is due to the first order dissociation
6.7 Irreversible Inhibition (kin6.m) 91

Ordered binding and obstucted dissociation (kin5b.m)

2
I0 = 1 mM
I0 = 10 mM
I0=100 mM
1.8

1.6
Bound ligand (µM)

1.4

1.2

0.8
0 0.2 0.4 0.6 0.8 1
time (sec)

Fig. 6.9 Ordered binding and obstructed dissociation monitored in the sub-second range. Reac-
tion scheme (6.18) is calculated with KD1 ¼ KD2 ¼ KD3 ¼ KI1 ¼ KI2 ¼ KI3 ¼ 10 mM, k1 ¼
ki1 ¼ k2 ¼ ki2 ¼ 1 mM1 s1. The initial concentrations are L0 ¼ 100 mM, R0 ¼ 1 mM. The
concentrations of the chasing ligand are 1,000 (+), 10,000 (x) and 100,000 (*) mM

of L from LLR. Compare Figs. 6.9 to 6.6 in this respect. Fast kinetics, i.e. reactions
which proceed in the sub-second time range, should be monitored with a stopped-
flow apparatus. Otherwise they would escape detection.

6.7 Irreversible Inhibition (kin6.m)

The final example of this chapter deals with irreversible reactions. Most reactions in
biochemistry are reversible, and even covalent bonds typically result from a series
of enzyme catalyzed reversible reactions. Nevertheless, proteins contain quite a few
reactive groups and reactive chemical compounds may bind irreversibly. The
kinetics of irreversible reactions is quite easy to understand. It becomes interesting
when irreversible binding leads to unexpected results. This section calculates the
influence of irreversible inhibition on dose–response curves of drug screening
programs.
Let us first discuss one simple chase experiment with an irreversible inhibitor.
Assuming that the inhibitor binds to the ligand binding site, it can only bind to the
free receptor R and not to the occupied receptor LR. The mechanism is shown in
reaction scheme (6.19). It is the same as reaction scheme (6.15) with ki1 ¼ 0.
92 6 Binding Kinetics

ð6:19Þ

Therefore, the function kin3F is used for the calculation of reaction scheme
(6.19) with kim1 ¼ 0 in line 22 of the main program kin6.m. The program
kin6.m is very similar to kin3.m, in particular since the initial equilibrium for L
and R can be calculated from a reversible reaction with fsolve and the function
EQ1F.m in line 32.

Figure 6.10 shows the decrease of the concentration of bound ligand after the
addition of different concentrations of inhibitor. The lowest concentration (1 mM)
does not significantly reduce the concentration of bound ligand within 10,000 s,
which correspond to almost 3 h. The highest concentration (10 mM) completely
inhibits the receptor within 20 min (1,200 s). This is to be expected, since irrevers-
ible reactions are of the second order and their rate in (6.19) is proportional to the
inhibitor concentration. At infinite times, all irreversible inhibitors at any concen-
tration would lead to complete inhibition, of course.
Now, how will the concentration dependence of irreversible inhibition influence
dose–response curves in drug screening campaigns? Drug screening procedures
typically are high throughput processes which are fully automated: The drugs in
6.7 Irreversible Inhibition (kin6.m) 93

Dissociation kinetics caused by irreversible inhibition (kin6.m)

1

0.8
Bound ligand LR (µM)

0.6

0.4

0.2

0
0 2000 4000 6000 8000 10000
time (sec)

Fig. 6.10 Dissociation kinetics caused by irreversible inhibition. Reaction scheme (6.19) is
calculated with KD1 ¼ 10 mM, k1 ¼ 0.001 mM1 s1 and ki1 ¼ 0.00001 mM1 s1. The initial
concentrations are L0 ¼ 100 mM, R0 ¼ 1 mM. Thirteen inhibitor concentrations (one solid line
each) are logarithmically distributed ranging from 1 to 10,000 mM

question are incubated for a fixed time (in our example 10 min) with the drug target
(typically an enzyme, but let us call it a receptor R in this chapter). Then the assay is
started by adding substrate (or ligand L). The assay may run another 10 min and the
activity (proportional to LR) is recorded after this time. Such a protocol can be
reproduced with a computer program, and kin6b.m is a simple example for this.
The core of the program kin6b.m is contained within two loops in lines 32–45.
94 6 Binding Kinetics

The initial concentrations (incubation of receptor R with inhibitor I) are defined

in lines 28–32, with the exception of the initial inhibitor concentration I1 which is
set in line 33 within the loop. The loop calculates the irreversible binding of I to R
within the incubation time of 10 min (600 s, as defined in line 25). The differential
equations are solved in line 35 and the resulting concentrations for R, I and IR are
used as starting concentrations R2 ¼ R, I2 ¼ I, IR2 ¼ IR for the assay. Since
these starting concentrations have been computed from the second time point of the
Matrix M, the results of M are directly assigned to the second vector of initial
parameters x2 in lines 36–40. Each of these k column vectors correspond to one
inhibitor concentration I00(k). The result of the actual binding assay is the
concentration of complexes LR(k) for all inhibitor concentrations I00(k), as
given in line 44. The corresponding reversible binding (one site, adapted
R0 ¼ 0.91, KD ¼ 200 mM) is calculated from (3.1) in line 48. This reversible
binding is included in Fig. 6.11 as a reference.
Figure 6.11 shows a dose–response curve for irreversible inhibition which is
similar to the dose–response curves in Fig. 5.12, where the effect of multiple
allosteric inhibition is calculated. Again, the dose–response curve is not symmetric
and thus does not correspond exactly to a logistic curve. If the curve for irreversible
inhibition was fitted to a logistic function, the resulting Hill coefficient would larger
than one. As noted in Sect. 5.8, Hill coefficients larger than one are typical for high
References 95

High throughput screens with irreversible inhibitors (kin6b.m)

1
Irreversible Inhibition
Reversible Inhibition

0.8
Bound ligand (µM)

0.6

0.4

0.2

0
0 0.5 1 1.5 2 2.5 3 3.5 4
log (Inhibitor concentration)

Fig. 6.11 High throughput screens with irreversible inhibitors. Reaction scheme (6.12) is calcu-
lated with KD1 ¼ 10 mM, k1 ¼ 0.001 mM1 s1 and ki1 ¼ 0.00001 mM1 s1. The initial
concentrations are L0 ¼ 100 mM, R0 ¼ 1 mM. A typical screening protocol was calculated with
inhibitor incubation time of 10 min followed by an assay of 10 min. Activity competition curve for
an irreversible inhibitor (o) and reversible inhibition () is shown

throughput screening campaigns [6]. Therefore, one reason for larger Hill
coefficients may lie in irreversible interactions of reactive compounds to reactive
sites on a drug target. This is not unreasonable, since the compound libraries
employed in screening campaigns contain quite a few chemically active
compounds. Most likely there is more than one reactive group on a protein, so
that a combination of multiple allosteric interactions in conjunction with irrevers-
ible inhibition may explain any Hill coefficient larger than one.
How to modify the sample program. The irreversible reaction of Fig. 6.11 can be
modified to a reversible reaction simply by changing kim1 in line 22. It will
become obvious that not only irreversible reactions, but also slow reversible
reactions may lead to the type of dose–response curve shown in Fig. 6.11.

References

1. Hindmarsh AC (1983) ODEPACK, a systematized collection of ODE solvers. In: Stepleman

RS et al (eds) IMACS transactions on scientific computation, vol 1. Scientific Computing,
North-Holland, Amsterdam, pp 55–64
2. Gutfreund H (1995) Kinetics for the life sciences. Receptors, transmitters and catalysts.
Cambridge University press, Cambridge
96 6 Binding Kinetics

3. Prinz H, Striessnig J (1993) Ligand-induced accelerated dissociation of (+)-cis-diltiazem from

L-type Ca2+ channels is simply explained by competition for individual attachment points. J
Biol Chem 268:18580–18585
4. Klebe C, Prinz H, Wittinghofer A, Goody RS (1995) The kinetic mechanism of Ran-nucleotide
exchange catalyzed by RCC1. Biochemistry 34:12543–12552
5. Prinz H, Maelicke A (1983) Interaction of cholinergic ligands with the purified acetylcholine
receptor protein II. Kinetic studies. J Biol Chem 258:10273–10282
6. NIH Data base: http://www.ncbi.nlm.nih.gov/sites/entrez. Chem BioAssay
Chapter 7
Enzyme Kinetics

Any type of enzyme catalyzed reactions can be calculated from their differential
equations, but steady-state equilibrium hides most intrinsic reactions. Mathemati-
cally, this problem corresponds to stiff differential equations. Empirically, initial
velocities have been used to account for steady-state conditions. Competitive,
noncompetitive, uncompetitive, and cooperative inhibition is calculated from initial
velocities and steady-state equilibria. Substrate inhibition is calculated as progress
curves from differential equations. At the end of this chapter, the reader should be
able to calculate all feasible enzyme mechanisms and translate it to traditional
interpretations.

7.1 Progress Curves (enz1.m)

Enzyme kinetics can be calculated from the differential equations of a reaction

scheme, just like any other kinetic experiment. They differ from binding kinetics,
because enzymes are biocalalysts which usually are present in small amounts
relative to the concentrations of substrates and products. After an initial binding
phase, a steady-state condition is established, in which a constant rate of substrate is
turned into product. In most textbooks on enzyme kinetics, the steady state is
treated as a transient equilibrium, for which analytical solutions can be calculated
from Michaelis–Menten approximations (Sect. 3.1.4).
In the beginning of this chapter, we will calculate the full course of an enzymatic
reaction with all its concentrations of complexes, substrate, and product. The graph
which shows the concentration of product as a function of time is called a “progress
curve.” The initial part of this curve gives the initial velocity, which commonly is
used to calculate Michaelis–Menten enzyme kinetics (3.6). The simplest enzyme
mechanism describes the catalytic reaction of one substrate S to one product P.
When substrate binds to an enzyme E, it will form a complex ES. The enzyme may
then catalyze a reaction, which turns the substrate S into product P. Since this is

H. Prinz, Numerical Methods for the Life Scientist, 97

DOI 10.1007/978-3-642-20820-1_7, # Springer-Verlag Berlin Heidelberg 2011
98 7 Enzyme Kinetics

catalyzed by the enzyme, the product will initially be bound to the enzyme as EP.
Eventually it will dissociate to E and P:

ð7:1Þ

Reaction scheme (7.1) describes the simplest plausible scheme for a single
substrate – single product transformation. Note that the third reaction is read from
left to right and that the equilibrium dissociation constant for E and P therefore must
be KD3 ¼ k3/k3. Reaction scheme (7.1) is calculated in the subroutine function
enz1F.m.

These differential equations are readily solved with the Octave function lsode
in line 41 of enz1.m. For MATLAB, one has to select a solver capable of solving
stiff differential equations. These solvers have names ending with “s”. The solver
selected in line 41 of enz1M.m is ode23s. The MATLAB universal solver which
had been used for binding kinetics was ode45, which is too slow for steady-state
kinetics and its stiff differential equations with fast on and off reactions k1 and k1.
Figure 7.1 shows a series of progress curves calculated from reaction scheme
(7.1) and different rate constants for the initial step k1. The initial parts of the
progress curves give the initial velocities for product increase or substrate decrease.
With the exception of the two lowest rate constants, the initial velocities are the
same within experimental error. Only initial velocities are taken as experimental
data for Michaelis–Menten kinetics, and Fig. 7.1 confirms that these values indeed
are safe to measure.
Figure 7.1 in its printed form is a black and white illustration, but when the
program enz1.m is run in the octave window (or enz1M.m in MATLAB), two
series of colored graphs are shown. Each value for k1 has the same color, be it for
product formation (+) or substrate decrease (o). This is achieved with the variable
fill, a useful programming feature, described in lines 49 and 50:

P and S are matrixes of N time points and N2 different k1 values. t is a vector of

N time points. When P is plotted versus t without specifying a color, all N2
different curves will be displayed each with different colors. The plot command
has a repertoire of seven different colors for this. When the next set of data (t,S, in
this case) is called in the plot function, the numbering of the colors is continued, so
that after the first 8 colors, color 9, 10, 11. . ., identical to 2, 3, 4. . . are displayed. If
one wants to have the same set of colors displayed for the second set of data, one has
7.2 Progress Curves for a Weak Substrate (enz1b.m) 99

Progress curves for product increase and substrate decrease (enz1.m)

100

80
Concentration (µM)

60

40

20

0
0 500 1000 1500 2000 2500 3000 3500 4000
time (sec)

Fig. 7.1 Progress curves for product increase and substance decrease. Product increase (+) and
substrate decrase (o) is calculated from reaction scheme (7.1) with various association rate
constants. KD1 ¼ KD2 ¼ 100 mM, k2 ¼ 1 s1, k2 ¼ 0.001 s1, k3 ¼ 0.005 s1. The values
for k1 are 0.001. 0.0032, 0.01, 0.032, 0.1, 0.32, and 1 mM1 s1

to include a set of dummy data, which would fill colors 9, 10,. . .14 (a multiple of 7).
Then, the next set is displayed beginning with color 1 again. fill in line 49
therefore defines a matrix of dummy data. The previous number of plotted curves
N2 has to be subtracted from a multiple of 7 (14, in this case). This trick is
particularly useful when theoretical curves are fitted to experimental data (Fig. 8.8).
How to modify the sample program. The equilibrium dissociation constants KD1
and KD3 in enz1.m were set to 100 mM, for substrate and product alike. Therefore,
the final equilibrium of substrate and product is determined by KD2, which is the
same as k1/k2. When this parameter in line 21 is set to 1, both, substrate and
product concentrations will approach 50 mM at infinite times. Enzymes are
catalysts, which will catalyze both, the transition from substrate to product and
from product to substrate. Reaction scheme (7.1) accounts for it.

7.2 Progress Curves for a Weak Substrate (enz1b.m)

Most enzyme kinetics show progress curves similar to Fig. 7.1. We had performed
enzyme kinetics with more than ten different protein phosphatases in a drug-
screening campaign. As expected, most of the progress curves looked similar to
100 7 Enzyme Kinetics

Fig. 7.1. In some cases, however, the progress curves looked different (Fig. 8.3).
The endogenous substrates of protein phosphatases are phosphorylated proteins,
which are almost impossible to obtain in a defined state of phosphorylation.
p-nitrophenyl phosphate (pNPP) is a common substitute. Its affinity is orders of
magnitude lower than the endogenous substrates, but its phenolate product can be
monitored at 405 nm. For some phosphatases like CDC25, its affinity is extremely
low with a Km value around 100 mM. Moreover, its maximal reaction rate also was
low, so that high enzyme concentrations were required in order to observe the
reaction. In this case, the progress curves did not resemble Fig. 7.1 and instead
exhibited a pronounced lag phase (Fig. 8.3).
Phosphatase catalyzes a one product, two substrates reaction. We will not use
this enzyme as an additional example, but modify the rate constants of scheme
(7.1), in order to understand under which conditions such a lag phase in an
enzymatic progress curve may appear. One usually would apply a fitting routine
such as described in Chap. 8 in order to find suitable sets of rate constants.
Instead, we will vary the parameters (rate constants and concentrations), until a
reasonable explanation for this lag phase is found. Such a task is made easier when
all relevant parameters are shown explicitly in the final graph, and the command
text (x,y,’label’,p1,v1) is used eight times in lines 51–66. The first two
numbers of text are x and y coordinates, where the written text should appear.
'label' is a string of characters which is written at the x, y location. This can be
followed by one or more property-value pairs (p, v). When the property value pair
p1 ¼ 'units' and v1 ¼ 'normalized' is used, the x and y coordinates are
both normalized to the range between 0 and 1.0. The string ’label’ may be
written as a row vector of strings, for example: [ 'KD1 ¼ ',num2str(KD1),'
mM']. The first and third vector elements are simple strings ' KD1 ¼ 'and ' mM'.
The second element is an interesting function: num2str(number) converts any
number to a string. Replacing number with a parameter, such as KD1, allows
octave to convert it into a string and to show it in the current plot with the text
command.

Indeed, one can obtain an initial sigmoid increase of product formation [(+) in
Fig. 7.2]. Of course, there may be different reaction schemes or different sets of
parameters which give similar curves, but the message is clear: Substrates with an
extremely small catalytic rate constant k2 may need a relatively long time to reach
steady state.
Figure 7.2 shows a lag phase correlated to the production of EP (x) from ES (o)
in reaction scheme (7.1). Such observations cannot be explained with classical
Michaelis–Menten assumptions (see Sect. 3.1.4). This example is presented here
because initial velocity enzyme kinetics generally look simple, and their underlying
assumptions are valid and allow for easy data analysis. But in some cases one has to
go back and calculate the full enzymatic process without these assumptions.
7.3 Michaelis–Menten Enzyme Kinetics (enz2.m) 101

Progress curve and concentrations of

complexes for a weak substrate (enz1b.m)
2
P
ES k1 = 0.01 µM-1s-1
EP KD1 = 100 µM
k2 = 0.002 s-1
1.5 KD2 = 0.001
k3 = 0.001 s-1
Concentration (µM)

KD3 = 100 µM
S0 = 100 µM
E0 = 1 µM
1

0.5

0
0 500 1000 1500 2000 2500 3000 3500 4000
time (sec)

Fig. 7.2 Progress curve and concentrations of complexes for a weak substrate. Reaction scheme
(7.1) is calculated with enz1b.m and the parameters given in the figure itself

How to modify the sample program. The program kin1b.m has been designed
to vary all parameters of scheme (7.1). These parameters are entered in lines
18–30. Do it!

7.3 Michaelis–Menten Enzyme Kinetics (enz2.m)

Michaelis–Menten kinetics is a reasonable simplification of reaction scheme (7.1).

One assumes a rapid initial equilibrium between free and bound substrate. From
bound substrate ES the product is released as a time-limiting (slowest) step. The
first-order rate constant of product release is named kcat.

ð3:4Þ

Scheme (3.4) corresponds to Michaelis–Menten kinetics (section 3.1.4) and the

differential equations are part of the subroutine function enz2F.m
102 7 Enzyme Kinetics

Progress curve for Michaelis-Menten kinetics (enz2.m)

100
P

80
Concentration of product (µM)

60

40 k1 = 10 µM-1s-1
KD = 20 µM
kc = 0.05 s-1
S0 = 100 µM
20 E0 = 1 µM

0
0 500 1000 1500 2000 2500 3000 3500 4000
time (sec)

Fig. 7.3 Progress curve for Michaelis–Menten kinetics. Reaction scheme (3.4) is calculated with
enz2.m and the parameters shown in the figure

Of course, the results depend on the set of parameters. Of these, only KD1 (k1/k1)
and kcat are significant, when a fast association rate constant k1 is chosen
in accordance with the Michaelis–Menten assumptions. This is done here with
k1 ¼ 107 M1 s1. When this k1 is changed by a factor of 10, it does not affect the
progress curve shown in Fig. 7.3. When these assumptions are not made, the full set
of differential equations in enz2.m allows the calculation of progress curves with
all feasible rate constants of the reaction scheme (3.4).
Figure 7.3 shows a progress curve for an irreversible transition from substrate S
to product P [scheme (7.2)]. Therefore, at infinite times the substrate concentration
always will be zero and the product concentration will be the same as the initial
substrate concentration. The initial part of this curve is linear for a considerable
time range. This linear part is the only part which is required to determine the initial
velocity. The initial velocity is the experimental value used for a Lineweaver–Burk
plot (Fig. 3.6) and the corresponding data evaluation.

7.4 Lineweaver–Burk Plots from Progress Curves (enz3.m)

Comparing Figs. 7.3 to 7.1 shows that progress curves calculated from the
simplified reaction scheme (3.4) give similar results to those calculated from the
complex scheme (7.1). Michaelis–Menten kinetics usually is not evaluated by
means of progress curves, but from measurements of initial velocities at different
7.4 Lineweaver–Burk Plots from Progress Curves (enz3.m) 103

substrate concentrations. Under steady-state conditions, this corresponds to an

equilibrium-binding curve, as discussed in Sect. 3.1.3. We will therefore calculate
reaction scheme (7.1) at different substrate concentrations, determine the initial
velocity from this, plot it as a Lineweaver–Burk Plot and determine Km and kcat
from there. All this is done in the program enz3.m. The vector S00 of N2 substrate
concentrations is defined in line 31:

N2 gives the number of points. The first parameter in the function logspace is
the logarithm of the first number, the second parameter the logarithm of the last
number. The numbers therefore range from 1 (100) to 1,000 (103). They are used in
the main program within a loop:

Within the loop, each step calculates the differential equations of enz1F for an
association reaction of E0 with a different concentration S0. The resulting product
concentrations as a function of time are the fifth concentrations of the function
enz1F and therefore are returned as the fifth column of the matrix M from lsode
in line 37. They are assigned to a vector Px in line 38. The rate of product
formation v0 is calculated from DP/Dt in line 39. A matrix P of product
concentrations for all N2 different substrate concentrations k and all time points
N is written in line 40 within the loop. The whole matrix can then be plotted with
the simple plot command in line 43, and the result is shown in Fig. 7.4.
Figure 7.4 shows a calculated series of progress curves. The initial velocities
were calculated from the slopes of third and second points in line 39. When these
values are plotted versus the logarithm of the substrate concentrations in Fig. 7.5,
the dose–activity curve follows the expected (sigmoid and symmetric) pattern of a
one-site binding curve.
Figure 7.5 shows a dose–activity curve which would be ideally suited for the
calculation of Km and kcat, from a least squares fit. The data points are equally
distributed and no bias is introduced by a transformation artifact. For this task,
numerical methods for nonlinear data fitting are described in Chap. 8.
The traditional way to analyze simple hyperbolic binding curves is a linear
regression in a linearized plot, such as in a double reciprocal plot. For enzyme
kinetics, this is the Lineweaver–Burk plot (Sect. 3.1.4), where the reciprocal initial
104 7 Enzyme Kinetics

Progress curves at different substrate concentrations (enz3.m)

20

15
Product Concentration (µM)

10

0
0 500 1000 1500 2000 2500 3000 3500 4000
time (sec)

Fig. 7.4 Progress curves at different substrate concentrations. Product increase (+) is calculated
from reaction scheme (7.1) and shown as a function of time. KD1 ¼ KD3 ¼ 100 mM, k2 ¼ 1 s1,
k1 ¼ 0.01 s1, k2 ¼ 0.001 s1, k3 ¼ 0.05 s1. The substrate concentrations were taken from a
logarithmic distribution between 1 mM and 1 mM

Dose-Activity curve (enz3.m)

5

4
Initial velocity v0 (nM/sec)

0
-1 -0.5 0 0.5 1 1.5 2 2.5 3
log ( Substrate concentration (µM))

Fig. 7.5 Logarithmic dose–activity curve from initial velocities determined from Fig. 7.4. KD1
¼ KD3 ¼ 100 mM, k2 ¼ 1 s1, k1 ¼ 0.01 s1, k2 ¼ 0.001 s1, k3 ¼ 0.05 s1. The initial
velocities were calculated from the third and second time point and multiplied by a factor of 1,000
7.4 Lineweaver–Burk Plots from Progress Curves (enz3.m) 105

velocity is plotted versus the reciprocal substrate concentration. Anyone with a

ruler can draw a straight line through such experimental data and extract the results
directly from the plot.
Computers use simple linear regression [1], which is an analytical method to fit a
straight line through any set of data points. Mathematically, it can be derived by
minimizing the residual sum of squares of a fit to this line. Unlike numerical
methods of nonlinear regression described in Sect. 8.1, it does not require an initial
estimate. The function linreg.m takes a pair of row vectors x and y and fits a
straight line through the data pairs x1/y1, x2/y2, x3/y3 . . . The result is a row vector
a with two components a(1) and a(2), for slope and intersect, so that the resulting
linear function is y ¼ a(1)*x + a(2).

Note line 4, which is a function of a function in order to determine the number of

points. size(x) returns a row vector of the number of rows and columns of x.
max(size)returns the maximal value of size, which is the number of points. Lines
5–7 give the analytical solution for slope a(1) and intersect a(2) [1]. sum(x) returns
the sum of all elements of x. x*y' in line 5 is a Scalar product of vectors, which is
equal to the sum of xi*yi. This is the only example in this textbook, where a scalar
product of vectors is used instead of element by element multiplication of vectors.
Simple linear regression requires double reciprocal transformation of data.
Reciprocals for the Lineweaver–Burk plot are calculated in lines 58 and 59 of
enz1.m. The resulting vectors rS (for reciprocal substrate concentrations) and rv
(for reciprocal velocities) are then fitted to a straight line with the help of linreg
(line 60). From this, Km, vmax, and kcat are calculated in lines 61–63. The
reciprocal data are plotted in line 64, and the result is shown in Fig. 7.6. The
results of the linear regression is the straight line y ¼ a(1)*rS + a(2). When this is
included in the plot command of line 64, the line only covers the range of the
experimental data. For a Lineweaver–Burk plot, one wants to elongate the line until
its intersection with the x-axis. For this, the command line in line 69 was used. It
plots a straight line from a first data point to a second data point. The coordinates for
these data points are defined in line 65 and 66. Likewise, the y-axis was drawn with
the command line(x1,y1) in line 70. A Lineweaver–Burk plot is not particu-
larly suited for showing the whole significant concentration range, since it tends
to magnify the low concentrations inappropriately. Therefore, the data for the
lowest three substrate concentrations were excluded. The command axis in line
71 takes the range from the intersection of the linear regression with the x-axis (the
Km-value) and 110% of the reciprocal value of the fourth substrate concentration
for the x-axis. Likewise, the scale for the y-axis was chosen between zero and 110%
of the fourth reciprocal initial velocity.
106 7 Enzyme Kinetics

Lineweaver-Burk Plot (enz3.m)

3000
linear regression
Km = 10.857 µM
2500 kcat = 0.048968 s-1

2000
1 / Initial velocity

1500
Parameters scheme 6.1
k1 = 0.01 µM-1s-1
1000 KD1 = 100 µM
k2 = 1 s-1
KD2 = 0.001
k3 = 0.05 s-1
500 KD3 = 100 µM
E0 = 0.1 µM

0
0 0.2 0.4 0.6 0.8 1 1.2
1 / Substrate concentration

Fig. 7.6 Lineweaver–Burk plot from reaction scheme (7.1). The rate constants from Figs. 7.4 and
7.5 printed in the lower right corner were used to calculate reaction scheme (7.1). The initial slope
of the progress curves obtained at different substrate concentrations was plotted versus the
substrate concentrations in a double reciprocal manner. From this, Km and kcat were calculated
from a linear regression and printed in the upper left corner

Figure 7.6 shows the Lineweaver–Burk plot for the progress curves in Fig. 7.4
and the dose–response curve in Fig. 7.5. The fitted parameters Km and kcat corre-
spond to reaction scheme (3.4).
How to modify the sample program. The program enz3.m has been designed to
compare reaction scheme (7.1) with the classical evaluation of Lineweaver–Burk
plots. It may be interesting to vary the parameters in lines 18–25, in order to
understand the properties of the Michaelis–Menten approximation. The same
program can be used as a template whenever the question arises how any reaction
7.5 Substrate Inhibition (enz4.m) 107

scheme would appear when analyzed as Michaelis–Menten kinetics. In this case,

the differential equations of the new reaction scheme would have to be written as a
subroutine similar to enz1F.m, and the name of this subroutine would have to be
called with lsode of line 37.

7.5 Substrate Inhibition (enz4.m)

Every enzyme is different and many enzyme mechanisms are discussed in the
literature, so that there are many candidates for suitable examples when we want
to exploit the power of numerical methods. Some enzymes are inhibited by their
own substrates [2, 3], and their initial velocities decrease at high substrate
concentrations. The corresponding Lineweaver–Burk plots are not linear any
more. There are numerous reaction schemes accounting for such an observation.
Some of these have been calculated for acetylcholinesterase [2, 4, 5], an enzyme (E)
which hydrolyzes acetylcholine (S) and produces choline (P1) and acetic acid (P2).
The acetylated enzyme (EP2) has been identified as an intermediate, and substrate
binding to the acetylated enzyme has been suggested as the key step of substrate
inhibition leading to the inactive EP2S. This corresponds to reaction scheme (7.2).
Note that inhibition is caused by a ternary complex of substrate and product P2.
Therefore, scheme (7.2) also had been referred to as “product inhibition” [5].

ð7:2Þ

The dissociation of P2 from EP2S (the deacetylation in the case of acetylcholin-

esterase), giving the complex ES is an additional step which has been suggested [4]
in order to explain some characteristics of the experimental data. The differential
equations for scheme (7.3) are the main part of the function enz4F.m:
108 7 Enzyme Kinetics

Progress curves for substrate inhibition (enz4.m)

12

10
Concentration of product (µM)

0
0 5000 10000 15000 20000 25000 30000 35000 40000
time (seconds)

Fig. 7.7 Progress curves for substrate inhibition (7.2). KD1 ¼ 100 mM, KD2 ¼ KD3 ¼ 100 mM.
k1 ¼ k4 ¼ 0.01 mM1 s1. k2 ¼ 1 s1, k3 ¼ 0.05 s1, k5 ¼ 0.001 s1. The enzyme concentration
is 10 nM. The substrate concentration varied in a logarithmic scale from 1 mM to 1 mM

The complex EP2 is involved in reactions 2, 3, and 4. Line 14 therefore became

too long and was extended with the continuation marker (. . .). The main program
enz4.m calls the function enz4F in line 47 and displays the results in four different
plots. First, the direct progress curves are shown in Fig. 7.7.
Figure 7.7 shows progress curves calculated with the program enz4.m and
plotted in line 53. There are 13 curves for 13 different substrate concentrations
ranging from 1 mM to 1 mM in a logarithmic scale. The lowest curve is the product
formation after the addition if 1 mM substrate. Note its curvature, because the low
substrate concentrations sustain the steady-state equilibrium only for a relatively
short time. The other curved lines show substrate concentrations of 1.8, 3.2, 5.6, 10,
and 18 mM, until a maximum is reached for a substrate concentration of 32 mM.
From there on, increasing substrate concentration leads to a decrease in the slope,
but the curvature for these high substrate concentrations is significantly decreased.
The long linear part reflects stable steady-state equilibrium.
Rather than measuring and calculating progress curves, enzyme kinetics usually
is evaluated from the initial velocities which reflects the steady state o enzyme
catalysis. The initial velocities are calculated in the main program of enz4.m,
within the loop (44–51) for the N2 different substrate concentrations.
7.5 Substrate Inhibition (enz4.m) 109

The initial velocity v0 in line 49 is calculated from the difference of the third
and second time point. For experimental data, the first time point always should be
avoided because there may be experimental artifacts. Even for theoretical data, this
point should not be used for the calculation of the initial velocity, because the
steady state is not achieved at time t ¼ 0. The data are plotted with the second plot
command in line 60.
Figure 7.8 shows the initial velocity calculated from the progress curves in
Fig. 7.7. The substrate concentrations of Fig. 7.8 only cover the range from 1 to
100 mM, in order to show the significant characteristics in this representation. For
the first seven substrate concentrations, it shows an increase expected from a simple
110 7 Enzyme Kinetics

Direct Activity Plot for substrate inhibition (enz4.m)

2.5
Initial velocity (µM)

1.5

0.5

0
0 20 40 60 80 100
Substrate concentration (µM)

Fig. 7.8 Linear dose–activity plot for substrate inhibition (7.2). KD1 ¼ 100 mM, KD2 ¼ KD3
¼ 100 mM. k1 ¼ k4 ¼ 0.01 mM1 s1. k2 ¼ 1 s1, k3 ¼ 0.05 s1, k5 ¼ 0.001 s1. The enzyme
concentration is 10 nM. The substrate concentration varied in a logarithmic scale from 1 mM to
100 mM. The initial velocity v0 was calculated from the slope between the third and second time
point in Fig. 7.7

(hyperbolic) binding curve. For higher substrate concentrations the initial velocity
decreases. For real experiments, such a decrease might be masked by experimental
error and thus may escape notice. Here, of course, the reaction is calculated over a
broader concentration range as shown in Fig. 7.10.
The Lineweaver–Burk plot (Fig. 7.9, plot 3 in line 79 of enz4.m) of these data
shows a significant curvature for high substrate concentrations. Remember that the
low substrate concentrations in this representation are located in the upper right part
of the graph. The straight line is fitted to the six lowest concentrations with the
function linreg as explained in the previous section. The respective program code
of enz4.m is located between lines 68 and 78. Line 70 creates a new vector from a
range of values of a larger vector. A range is defined with a colon, so that 1:6 is the
range of the first six elements, and rS(1:6) is the range of vector elements between
rS(1) and rS(6). rS is the vector of reciprocal substrate concentrations, and rs1
therefore is a new vector consisting of the first six data points of rS.
Figure 7.10 (plot 4, line 79 of enz4.m) shows a logarithmic dose–activity
curve, whereby the activity of the enzyme is defined as the initial velocity of
product formation. These logarithmic representations all begin with a sigmoid
increase and end with a phase where a final value is gradually approached. They
are useful for visualizing a large range of experimental data, and they are particu-
larly suited to compare fitted curves to experiments (Figs. 8.5–8.7).
7.6 Enzyme Inhibition with Reversible Inhibitors (Enz5.m) 111

Lineweaver-Burk Plot for substrate inhibition (enz4.m)

25000

linear regression for low substrate

Km = 10.796 µM
kcat = 0.046952 s-1
20000
1 / Initial velocity

15000

Parameters scheme 6.3

k1 = 0.01 µM-1s-1
10000
KD1 = 100 µM
k2 = 1 s-1
KD2 = 100000
k3 = 0.05 s-1
5000 KD3 = 100000 µM
K4 = 0.01 µM-1s-1
KD4 = 100 µM
k5 = 0.001 s-1
E0 = 0.01 µM
0
0 0.2 0.4 0.6 0.8 1
1 / Substrate concentration

Fig. 7.9 Lineweaver–Burk plot for substrate inhibition (7.2). KD1 ¼ 100 mM, KD2 ¼ KD3
¼ 100 mM. k1 ¼ k4 ¼ 0.01 mM1 s1. k2 ¼ 1 s1, k3 ¼ 0.05 s1, k5 ¼ 0.001 s1. The enzyme
concentration is 10 nM. The substrate concentration is varied in a logarithmic scale from 1 mM to
100 mM. The initial velocity v0 is calculated from the slope between the third and second time
point in Fig. 7.7. The parameters indicated in the lower right corner are used to calculate reaction
scheme (7.3). The initial slope of the progress curves obtained at different substrate concentrations
is plotted versus the substrate concentrations in a double reciprocal scale. From this, Km and kcat
were calculated in enz4.m from a linear regression of the six lowest substrate concentrations

7.6 Enzyme Inhibition with Reversible Inhibitors (Enz5.m)

Probably most enzymatic assays today are performed in high throughput screening
facilities. These assays are performed at very low enzyme and relatively low
substrate concentrations, so that the resulting slow reaction times agree with the
pipetting processes employed. Substrate inhibition or other variations of the
Michaelis–Menten kinetics usually are not observed under these conditions, so
that steady-state assumptions are valid. Binding of substrate S to the enzyme E
leads to an active enzyme substrate complex ES. The activity of this complex often
is measured by a spectroscopic assay which in turn may involve additional
components. Adding inhibitors 10–20 min before to the addition of substrate
usually establishes equilibria between enzyme and inhibitor. A few minutes after
addition of substrate steady-state equilibria between substrate and all complexes is
established. Assuming a maximum of two inhibitor binding sites on the enzyme,
one may use scheme (7.3) to calculate the coupled equilibria. Note that the reaction
112 7 Enzyme Kinetics

Logarithmic Activity Plot for substrate inhibition (enz4.m)

3

2.5
Initial velocity (µM/sec)

1.5

0.5

0
0 0.5 1 1.5 2 2.5 3
log (Substrate concentration (µM))

Fig. 7.10 Logarithmic dose–activity plot for substrate inhibition (7.2). KD1 ¼ 100 mM, KD2 ¼
KD3 ¼ 100 mM. k1 ¼ k4 ¼ 0.01 mM1 s1. k2 ¼ 1 s1, k3 ¼ 0.05 s1, k5 ¼ 0.001 s1. The
enzyme concentration is 10 nM. The substrate concentration is varied in a logarithmic scale from
1 mM to 100 mM. The initial velocity v0 is calculated from the slope between the third and second
time point in Fig. 7.7

scheme (7.3) is a sequential scheme like reaction scheme (2.18). It is useful for
equilibrium studies, where individual values for equilibrium dissociation constants
of equivalent intrinsic sites can be calculated from (2.30). For binding kinetics,
scheme (7.3) should not be used without reason.

ð7:3Þ

ES is the active complex, the concentration of which is proportional to the

enzymatic activity. For all practical cases, its equilibrium dissociation constant is
the same as the Michaelis constant Km. The different KI are the equilibrium
dissociation constants of the inhibitor and the different complexes. The enzymatic
activity is proportional to ES, but active ESI or ESI2 cannot be ruled out. The
7.6 Enzyme Inhibition with Reversible Inhibitors (Enz5.m) 113

concentrations of all complexes can be calculated from the concentrations of the

free concentrations [S], [E], and [I]. Just as discussed in Chap. 5, this gives three
nonlinear equations for E0, S0, and I0 as a sum of free and bound enzyme, substrate
and inhibitor, respectively. These equations are solved with fsolve in line 41 of
the program Enz5.m. The function Enz5F.m contains the equations

Traditionally, enzyme inhibition curves are presented as series of

Lineweaver–Burk plots. For the case of one inhibitor binding site they have been
calculated with analytical methods [6–8]. When the corresponding linear
regressions intersect on the Y-axis (Fig. 7.12), the inhibitor is competitive, when
they intersect on the X-axis (Fig. 7.13), the inhibitor is named noncompetitive, and
when the Lineweaver–Burk plots run parallel, the inhibition mechanism was named

Enzyme inhibition with two inhibitor sites (Enz5.m)

2000
1 / Concentration of active ES

1500

1000

500

0
-0.1 -0.05 0 0.05 0.1 0.15 0.2
1 / Substrate Concentration

Fig. 7.11 Lineweaver–Burk plot calculated from reaction scheme (7.3). Km was taken as 10 mM,
and all KI values with the exception of KI3 were taken as 100 mM. KI3 was taken as 10 mM. E0
was 10 nM. The inhibitor concentrations were 0, 25, 50, 75, and 100 mM
114 7 Enzyme Kinetics

to be uncompetitive. In many cases we found none of these, and sometimes the lines
did not even intersect in one point, as shown in Fig. 7.11.
Figure 7.11 shows a calculated set of data which corresponds to some of our
experimental findings. For each inhibitor concentration, the reciprocal initial
velocities lie on a straight line, but these lines need not intersect when there is
more than one inhibitor binding site.
The program Enz5.m is useful for the calculation of a many different inhibition
patterns in enzyme kinetics. It calculates steady-state equilibria, and therefore
corresponds to the usual approach to enzyme kinetics [6]. It allows the calculation
of many limiting cases, just by the manipulation of the equilibrium dissociation
constants. For example, if one wants to calculate binding only to one site, the
equilibrium dissociation constants for the second site (KI3 and KI4) can be set
many orders of magnitude higher than the other equilibrium dissociation constants.
The simple case of competitive binding at one site is written in Enz5.m
with Octave code as: KM ¼ 10; KI1 ¼ 10; KI2 ¼ 10000000;
KI3 ¼ 10000000; KI4 ¼ 10000000;
Figure 7.12 indeed shows the characteristic features of competitive binding.
When a substrate competes with the inhibitor, the extrapolated value at infinite
substrate concentrations must be identical at all inhibitor concentrations. Infinitely
high substrate concentrations correspond to 1/S ¼ 0, and all lines corresponding to
all inhibitor concentrations have to intersect at vmax, the maximal initial velocity.
The intersection with the x-axis equals 1/Km, so that the affinity is decreased (The
apparent Km is increased) with increasing inhibitor concentration.
The next classical example concerns noncompetitive binding. In this case, the
inhibitor may bind to the free enzyme E and to the occupied enzyme ES. When the
affinity for both is the same, then the binding of inhibitor cannot change the affinity
for the substrate.
Figure 7.13 shows noncompetitive inhibition. The lines intersect at 1/Km on
the x-axis. The intersection at the x-axis gives the Km value, so that the affinities for
the substrate are the same at all inhibitor concentrations. The substrate cannot
compete with the inhibitor, since the ternary complex ESI is stabilized with
increasing S at all inhibitor concentrations. Therefore, the lines do not intersect at
the same vmax value.
Incidentally, the theoretical lines in Enz5.m are calculated slightly different
from the theoretical lines in enz3.m. In both cases, the function linreg.m is
used, but whereas in enz3.m the lines are plotted separately from the plot with the
command line, they are now included as theoretical lines in the plot command.
This is done in 48–54 of the program Enz5.m:
7.6 Enzyme Inhibition with Reversible Inhibitors (Enz5.m) 115

Competitive inhibition (Enz5b.m)

2000
1 / Concentration of active ES

1500

1000

500

0
-0.1 -0.05 0 0.05 0.1 0.15 0.2
1 / Substrate Concentration

Fig. 7.12 Lineweaver–Burk plot for competitive inhibition (Enz5b.m). Reaction scheme (7.3),
Km ¼ 10 mM, KI1 ¼ 10 mM, KI2 ¼ KI3 ¼ KI4 ¼ 10 M, E0 was 10 nM. The inhibitor
concentrations were 0, 25, 50, 75, and 100 mM

Noncompetitive inhibition (Enz5c.m)

2000
1 / Concentration of active ES

1500

1000

500

0
-0.1 -0.05 0 0.05 0.1 0.15 0.2
1 / Substrate Concentration

Fig. 7.13 Lineweaver–Burk plot for noncompetitive inhibition (Enz5c.m). Reaction scheme
(7.3), Km ¼ 10 mM, KI1 ¼ KI2 ¼ 10 mM, KI3 ¼ KI4 ¼ 10 M, E0 was 10 nM. The inhibitor
concentrations were 0, 25, 50, 75, and 100 mM
116 7 Enzyme Kinetics

VS0 is the row vector of total substrate concentrations, as before, so that rS0
becomes the row vector of reciprocal substrate concentrations (line 48). VES is a
matrix, a row for inhibitor concentrations of column vectors at different substrate
concentrations. rES therefore is a corresponding matrix. Since linreg can only
compute single arrays (vectors), the columns for each inhibitor concentration k
have to be extracted from that matrix with rES(:,k)in line 52. The function
linreg requires row vectors as arguments, so that the column vector rES(:,k)
has to be transposed to a row vector with the transpose operator (') in line 52.
The resulting linear function is calculated in line 53. Note that the range for the
independent variable has been extended in line 50 to include values up to the
negative 1/KM.
Uncompetitive inhibition (Fig. 7.14) is observed when the inhibitor only binds to
the complex ES formed between the enzyme and the substrate. The inhibitor thus
increases the affinity (decreases the Km) and decreases the maximal velocity. The
increase in affinity is easy to calculate, but difficult to understand. One would
assume that any inhibitor should decrease the apparent affinity of a substrate. Here,
however, the inhibitor does not compete with the substrate for the free enzyme, and
in fact stabilizes the ternary complex ESI.
Most enzyme kinetics is measured during drug screening campaigns [9]. In this
case, the data are not evaluated as Lineweaver–Burk plots, but as dose–response

Uncompetitive inhibition (Enz5d.m)

1400

1200
1 / Concentration of active ES

1000

800

600

400

200

0
-0.1 -0.05 0 0.05 0.1 0.15 0.2
1 / Substrate Concentration

Fig. 7.14 Lineweaver–Burk plot for uncompetitive inhibition (Enz5c.m). Reaction scheme
(7.3), Km ¼ 10 mM, KI2 ¼ 10 mM, KI1 ¼ KI3 ¼ KI4 ¼ 10 M, E0 was 10 nM. The inhibitor
concentrations were 0, 25, 50, 75, and 100 mM
7.6 Enzyme Inhibition with Reversible Inhibitors (Enz5.m) 117

Four Inhibition-Activity curves(Enz5e.m)

5
Noncompetitive
Competitive
Uncompetitive
Two sites Cooperative
4
Concentration of active ES (nM)

0
-1 -0.5 0 0.5 1 1.5 2 2.5 3
Concentration of Inhibitor (µM)

Fig. 7.15 Dose–activity curves for reversible inhibition (7.3). (+) Noncompetitive inhibition
KI1 ¼ KI2 ¼ 20 mM. (x) Competitive (KI1 ¼ 10 mM), (o) uncompetitive (KI2 ¼ 10 mM) inhibi-
tion. (*) Cooperative inhibition at two sites (KI1 ¼ KI2 ¼ 10 mM, KI3 ¼ Ki4 ¼ 1 mM). E0 ¼ 10
nM, S0 ¼ 100 mM, KM ¼ 10 mM

plots, or dose–activity plots, or dose–inhibition plots. This corresponds to the calcu-

lation of active [ES] plotted versus the logarithm of the inhibitor concentration.
Figure 7.15 shows dose–activity curves for competitive inhibition (x), noncom-
petitive inhibition (+), and uncompetitive inhibition (o) for a single site as shown
in Figs. 7.12–7.14. These three dose–activity curves are identical and correspond
to a Hill coefficient of 1. Only when a second site becomes available, the shape of
the dose–response curve changes, and the slope at its inflection point becomes
steeper.
One note for MATLAB users: In the MATLAB environment, the solver
fsolve in the program Enz5e.m could not compute all values of Fig. 7.15, at
least in the MATLAB version R2009a. The highest inhibitor concentrations
lead to a numerical artifact as had been discussed in Sect. 2.6. Therefore, the
MATLAB program Enz5eM.m does not cover the whole range of inhibitor
concentrations.
Numerical artifacts generally have to be anticipated whenever numerical
methods are employed. When computed data show a continuous function with
one or two outliers, one can safely assume that these outliers have been generated
by numerical artifacts.
118 7 Enzyme Kinetics

References

1. http://en.wikipedia.org/wiki/Simple_linear_regression
2. Reed MC, Lieb A, Nijhout F (2010) The biological significance of substrate inhibition: a
mechanism with diverse functions. Bioessays 32:422–429
3. Sekulic N, Konrad M, Lavie A (2007) Structural mechanism for substrate inhibition of the
adenosine 5-phosphosulfate kinase domain of human 3-phosphoadenosine 5phosphosulfate
synthetase 1 and its ramifications for enzyme regulation. J Biol Chem 282:22112–22121
4. Hofer P, Fringeli UP (1981) Acetylcholinesterase kinetics. Biophys Struct Mech 8:45–59
5. Brockendahl H, M€ uller T-M, Verf€urth H (1968) Zur Kinetik der Produkthemmung. Hoppe
Seyler’s Z Physiol Chem 349:21–24
6. Cleland WW (1963) The kinetics of enzyme-catalyzed reactions with two or more substrates or
products II. Inhibition nomenclature and theory. Biochim Biophys Acta 67:173–185
7. Berg J, Tymoczko J, Stryer L (2002) Biochemistry. W. H. Freeman and Company, New York
8. Dixon M, Webb EC, Thorne CJR, Tipton KF (1979) Enzymes, 3rd edn. Longman, London,
p 126
9. Inglese J, Auld DS, Jadhav A, Johnson RL, Simeonov A, Yasgar A, Zheng W, Austin CP
(2006) Quantitative high-throughput screening: a titration-based approach that efficiently
identifies biological activities in large chemical libraries. Proc Natl Acad Sci USA
103:11473–11478
Chapter 8
Fitting the Data

Fitting models to experimental data is done by nonlinear regression routines, which

vary multiple parameters like rate constants, etc. until an optimal fit is achieved.
The resulting set of parameters need not be unique. Parameters often are correlated,
so that the variation of one parameter can be compensated by variations of other
parameters without reducing the quality of the fit. A correlation matrix helps to
clarify this point. Experimental procedures, strategies for the reduction of the
number of varied parameters, and global fits enhance the reliability of derived
rate constants. At the end of this chapter, the reader should be able to import data
from a spreadsheet, fit them to any reaction scheme and do a critical assessment of
the significance of the fitted values.

8.1 Multi-parameter Fits and Correlation of Parameters

(fit1.m)

A theory can only be verified on the basis of experimental data. When data and
theoretical curves are plotted together, both sets initially do not match, and
parameters have to be adjusted until a reasonable fit is achieved. The criterion, by
which the quality of such a fit is measured, is the sum of the squared differences
between experimental and theoretical points (8.1).

S(dati
theoi Þ2 (8.1)

The method to minimize the squares of these differences is commonly called the
method of “least squares”, so that the Octave routine is called leasqr and
MATLAB routine is lsqcurvefit. The actual function call in Octave and
MATLAB is different, but both functions give very similar results. They require
a vector x of independent variables, a vector dat of experimental data for each x,
and the name of the function which calculates the theoretical values theo as a

H. Prinz, Numerical Methods for the Life Scientist, 119

DOI 10.1007/978-3-642-20820-1_8, # Springer-Verlag Berlin Heidelberg 2011
120 8 Fitting the Data

function of x. Last but not least the routines require an initial set of parameters
(“pin”) which is modified until a good fit is achieved.
Unlike simple linear regression [1], calculated with linreg in Sect. 7.4,
parameters obtained from multi-parameter nonlinear regression need not be unique.
This is illustrated with a drastic example. The program fit1.m applies nonlinear
regression to a linear function containing an excess of parameters and compares
the result with simple linear regression. The problem of correlated parameters
immediately becomes obvious, so that one can deduce some general strategies for
multi-parameter data fitting.

8.1.1 The Sample Program fit1.m

The linear function (8.2) to be fitted as our drastic example contains an excessive
number of parameters.

y ¼ ða=bÞ  x þ c
d (8.2)

Function (8.2) gives a straight line with a slope of a/d and an interception of c–d
at the y-axis. The parameters to be fitted (a, b, c and d) are elements of a vector pin.
With this vector notation, the function (8.2) is translated to the function fit1F.m
called from leasqr in Octave.

For MATLAB, its function lsqcurvefit expects a reversed order of

arguments, so that the first line of the corresponding MATLAB function is

Everything else is the same for these functions, whether they are written in
Octave and called from leasqr or in MATLAB and called from lsqcurvefit.
The global statement (not used in the function fit1M.m) can be used to transfer
variables which are not varied in the nonlinear regression routine. The sample
program fit1.m calls the function leasqr in lines 22 and 23. Note that leasqr
expects column vectors for the arguments, so that the row vectors x and dat have
to be transposed (operator ’). Line 22 gives six results of leasqr, namely
fcurve, the resulting fitted theoretical curve, FP, the resulting fitted parameters
kvg and iter for information of the fitting process itself (not used here), corp,
the correlation matrix of parameters, and covp, the covariance matrix of
parameters.
8.1 Multi-parameter Fits and Correlation of Parameters (fit1.m) 121

A set of sample data is generated in lines 16–18. The function rand(1,N)

gives a row vector of N elements of random numbers between 0 and 1. Therefore,
(0.5-rand(1,N)) is a row vector with random numbers between
0.5 and 0.5,
and the result r in line 17 is a row vector with a mean value of 1 and a noise with
random numbers maximal 10%. The data set dat (line 18) therefore consists of a
noisy linear distribution as shown in Fig. 8.1.
Figure 8.1 shows that the nonlinear regression leasqr will find a solution, even
if the number of fitted parameters exceeds the number of meaningful parameters.

Nonlinear regression for a linear function (fit1.m)

1000

fitted values for y = (a/b)*x + c-d

a = –142.1
800 b = –14.493
c = 14.271
d = 4.8169
Calculated function y

Slope = 9.8038
600 Intercept = 9.4673

400

Initial Parameters for y = (a/b)*x + c-d

a = 10
200 b=1
c = 10
d=1

0
0 20 40 60 80 100
Independent variable x

Fig. 8.1 Multi-parameter fits of a linear function. The program fit1.m generates data (*) and
shows fitted linear functions (
). The results from simple linear regression linreg and from the
nonlinear regression leasqr overlap to one solid line
122 8 Fitting the Data

Both, y in line 20, the function fitted from linreg, and fcurve in line 22, the
function fitted with leasqr, give the same results. They are exactly superimposed
in Fig. 8.1, so that only one line is visible, although both functions are plotted with
the plot command in line 25.
Note the parameters returned from leasqr. They are shown as a, b, c, and d in
Fig. 8.1. Each time the sample program is executed (type fit1 in the Octave
terminal window), a different data set is generated and new parameters are fitted.
The resulting parameters a,b,c,d differ considerably for each execution of fit1.
Try it out! The erratic results for a,b,c,d are in contrast with the reasonable
fluctuations of pa(1) (¼Slope) and pa(2) (¼Intercept), returned from linreg.
Even though the results for a and b cannot be predicted, its quotient a/b is the
same as pa(1), the slope returned from linreg. Likewise, the difference c–d is
the same as pa(2), the intercept returned from linreg. All this is not surprising.
Exactly two parameters (slope and intercept) define a straight line and one cannot
obtain more than two parameters from a fit to such a line. But let us pretend that we
do not know this, and that we have performed numerous fits with the program fit1.
m, alas with the same data set. When we then plot the resulting parameter a as a
function of b, and the parameter c as a function of d, we find that all these data pairs
lie on straight lines:
a ¼ f(b) ¼ b  pa(1) (8.3)

c ¼ f(d) ¼ d þ pa(2) (8.4)

In the general case of multi-parameter fits to unknown functions, the parameters

which give the same (optimal) fit will not lie on straight lines, as in (8.3) and (8.4),
but will show a certain distribution. In the ideal case, this distribution is narrow, so
that a variation of one parameter x could not be compensated by the variation of
another parameter y, and still gives the same quality of a fit. The linear dependence
of two parameters can be described with the help of Pearson’s correlation coeffi-
cient [2–6]. This coefficient can have values between 1 and
1. If it is one, the
parameters are correlated, so that the increase in x can be completely compensated
by an increase in y. If it is
1, an increase in x can be compensated by a decrease in
y. If it is zero, the parameters are not correlated, and an increase in x cannot be
compensated by a variation of y. For a and b of (8.2) and (8.3), an increase in a can
be compensated by an increase in b. Its correlation coefficient is 1, the same as the
correlation coefficient of c and d in (8.4). For the ideal case of significant
parameters, a variation of x cannot be compensated by a variation of y and the
correlation coefficient should be near zero.
For MATLAB, the function lsqcurvefit returns no statistical information like
corp or covp. For this, one would have to use the function nlinfit from the
MATLAB Statistics toolbox. The MATLAB program fit1M.m differs from the
Octave program fit1.m only in lines 22 and 23.
8.1 Multi-parameter Fits and Correlation of Parameters (fit1.m) 123

The order of arguments in lsqcurvefit differs from leasqr, and this is

reflected in the order of arguments of fit1FM.m. The MATLAB function returns
the vector par of fitted parameters, but not the fitted theoretical curve. This can be
computed (line 23) from the same function (fit1FM) which was used for the
nonlinear regression in line 22. The MATLAB function lsqcurvefit accepts
row or column vectors, but size has to be identical for x and dat.

8.1.2 Strategies to Fit Correlated Parameters

fit1.m gives a drastic example to demonstrate the problem arising from the
correlation of parameters. For reaction schemes, one typical example of correlated
parameters are forward and back reaction rate constants under steady state
conditions. An increase of the forward rate constant of a reversible reaction
would be compensated by an increase of the back reaction rate constant, in order
to keep the equilibrium constant. Their correlation coefficient can be expected to be
1. For kinetic experiments it generally is a good idea not to vary both rate constants,
but to vary one rate constant and one equilibrium constant instead.
There generally are numerous rate constants or intrinsic equilibrium dissociation
constants which are important for the understanding of the mechanism, but which
cannot be determined experimentally. Reaction scheme (7.1), for example, is a
plausible minimal reaction scheme for enzyme kinetics, but it contains six rate cons-
tants, which simply cannot be fitted from progress curves. Progress curves are domi-
nated by steady state equilibrium (but see Sect. 7.2) so that quotients of rate constants
(i.e. the equilibrium constants) are fitted together with one of the rate constants.

ð7:1Þ

This still results in six parameters to be fitted, namely KD1, k1, KD2, k2, KD3,
and k3. With the exception of k3, all rate constants can be assumed to be fast, so that
they do not influence the shape of the progress curve. KD1 is the initial equilibrium
which corresponds to the Km value under some conditions. For reversible reactions,
one might assume the extreme case that the product has a similar affinity as the
substrate, so that KD3 ¼ KD1. When the reversible reaction (substrate formation
from product) is unlikely, one may set KD3 to a large value such as 1 M. This leaves
three parameters to be fitted, namely, KD1, KD2 and k3. Such a strategy does,
however, not prove the assumptions employed for the fitting routine, and a fit
performed under a set of assumptions does not return unambiguous experimental
results. Incidentally, scheme (3.4) of Michelis–Menten kinetics follows from such a
strategy and gives a commonly accepted simplified version of (7.1).
The next part of a fitting strategy consists of employing as many different
experiments as possible. Enzyme kinetics, for example, cannot be measured from
one progress curve. A series of curves has to be measured, and that series has to
124 8 Fitting the Data

cover a broad concentration range. When a series of experimental curves measured

under different conditions is fitted with one set of rate constants, the strategy is
named “global fit”. A global fit may generate additional parameters. For example,
there may be variations in the initial concentrations used in different experiments,
or there may be different background signals for different individual experiments.
As a rule, these additional parameters for individual experiments are not correlated
with the global ones.
Statistical values such as correlation coefficients and the sum of squares can help
to assess the quality of the results, but they never should be taken as standalone
criteria. There are at least three questions which have to be asked:
1. Does the fit reflect the properties of the data? If there is a systematic deviation
between experimental data and the fitted theoretical curve(s), either the model or
the set of parameters is not correct.
2. Can a variation of one parameter be compensated by the other ones? If, for
example, the parameter k1 resulting from a multi-parameter fit is changed by a
factor of 2 or 10 and kept constant, can a new fit with the other parameters
compensate this? A correlation coefficient will help to provide a strategy for the
choice of fixed or varied parameters.
3. How does the fit and how do the parameters vary with repetitions of the
experiments?
As will be discussed below, individual parameters cannot be determined unam-
biguously when they are correlated with others. But a successful fit proves that a
model with its derived set of parameters is adequate to quantify the fitted data.

8.2 Experimental Setup

Data analysis requires reliable data, and bioanalytical studies require the precise
handling of small volumes. Each pipetting step introduces an experimental error, so
that the first aim of experimental design must be the reduction of pipetting steps. A
second concern is the reproducibility. Many samples are prepared from living
organisms, so that they may vary from batch to batch. For a large quantitative
study one should never change the batch within a study. It is a better idea to pool
batches and to work from this pool. The same holds for all stock solutions.
Weighing inhibitors, substrates, buffers, etc. typically has a larger experimental
error than the subsequent pipetting steps.
Calibration errors can be avoided by dilution series, whereby a given stock
solution is diluted 1:1 by a series of dilutions with the same volume. This is
shown in Fig. 8.2. If the same pipette is used for all steps, there is no systematic
calibration error. Moreover, the concentrations will be distributed in a logarithmic
scale, so that experimental data can be equitably distributed over a broad range of
concentrations.
8.3 Entering the Experimental Data (fit2.m) 125

Fig. 8.2 Serial dilution 1:1. First, a volume V is dispersed to a series of tubes or microtiter wells.
Then the same volume V of a stock solution is added to the first tube with the same pipette to give
double the volume and half of the concentration. Then the same volume V is taken off with the
same pipette and added to the second well, and so forth. This procedure results in concentrations of
C0, C0/2, C0/4, C0/8 . . . The last volume V, taken from the last tube after the mixing step is
discarded

There are many other concerns for experiments intended for global fits. The
temperature for kinetic experiments has to be maintained; all controls have to be
performed at the same day, possibly before and after the experiment, etc. It only
requires common sense, but one should note that global fits may not work for older
experimental data which had not been recorded consistently.

8.3 Entering the Experimental Data (fit2.m)

Data input and output is done slightly differently in Octave and MATLAB.
MATLAB reads the Microsoft Excel format with the command xlsread. In
GNU Octave, dlmread only reads tab-delimited text files, which may be gene-
rated within Excel when a spreadsheet is saved in the *.txt format with the option
“tab separated”. The Octave function dlmread in line 13 of the sample program
fit2. m returns a numeric matrix M. The MATLAB function xlsread returns a
vector [M,C]with 2 elements, one numeric matrix M and one cell array C. The
argument GeniosPro_Sample.asc is the name of the spreadsheet to be read.
Here, it is the output of a microtiter plate reader. It consists of 78 readings of a
16  24 microtiter plate performed with a time difference of 27 s. These translate
into 384 kinetic experiments with 78 time points each. Additional information like
126 8 Fitting the Data

time and temperature becomes obvious when the sample data are opened as a
spreadsheet. Screening results from our Genios Pro reader are chosen as a realistic
example for experimental data.
Once the data are read in line 13, the size of the matrix M is determined in line 14.
For Octave, this size is [149,625], 1,496 rows and 25 columns. In MATLAB,
s ¼ [148,024]. Both languages obviously have a different algorithm to identify
numerical data. The readings are separated by three lines, so that there are
16 + 3 ¼ 19 lines for each experiment. Comparing M and GeniosPro_Sample.
asc, one can deduce the equation to calculate the number of time points in line 15.
From this, a three-dimensional matrix mes(i,k,j) is calculated from M in line 19.
The first dimension is time and the second and third dimensions are coordinates of the
microtiter plate. A vector time for N time points is constructed in line 20.

Reading data from a spreadsheet takes a while. When extracted data are used for
fitting procedures, it is a good idea to save them in a MATLAB (or Octave) format.
This is done in line 22 with the command save. The use of the corresponding
load command is shown in line 24. It restores all parameters, values, and variable
names after the clear command in line 23. The options and data formats for
save are different in Octave and MATLAB, but the syntax in lines 22 and 24 is
identical in both languages. The plot command in line 25 leads to Fig. 8.3, where 16
experiments of the first column of the microtiter plate are shown.
Figure 8.3 shows a screen for CDC25 inhibition with 16 substances from the first
column of a 384 well microtiter plate. The program fit2.m extracts data from a
worksheet and saves them in Octave format.
How to modify the program. The program fit2b.m retrieves these data from the
file save.mat, which had been saved in Octave format, and therefore is much faster.
When data are retrieved with the load command, all names are retrieved and the
variable names assigned in the program are overwritten. The command load
('save.mat'); therefore has to appear in the very beginning (line 12) of the
program fit2b.m. If it would appear later, the filename 'fit2b' (line 13) would be
overwritten with 'fit2', the filename which had been stored. The plot command in
line 14 plots all data. This is not possible in MATLAB, where three-dimensional
matrices cannot be plotted.
8.4 Fitting Substrate Inhibition 127

Screening with CDC25 (fit2.m)

0.9

0.8
Absorption 405 nm

0.7

0.6

0.5

0.4
0 500 1000 1500 2000
Time (seconds)

Fig. 8.3 Screening phosphatase activity of CDC25.The substrate pNPP was added to a concen-
tration of 50 mM, before the absorption at 405 nm was recorded. Data were retrieved with the
program fit2.m from the spreadsheet GeniosPro_Sample.asc, stored in Octave format as
sample.mat and retrieved with the load command

The resulting plot is not shown explicitly, since it is very similar to Fig. 8.3. Try
to plot different data with the program fit2b.m. For example, mes(:,3,:) is the third
row of the microtiter plate measured at all time points. Or (2:N,:,8) is the 8th
column of all but the first time points.

8.4 Fitting Substrate Inhibition

A fitting procedure can be a lengthy process. Once a solid set of experiments has
been translated into a data file and stored in Octave format, a fitting strategy has to
be devised. Usually this is accompanied with the development a suitable reaction
scheme. For substrate inhibition, reaction scheme (7.2) has been used successfully
[7]. It is an example for an enzyme mechanism which cannot be analyzed with
Michaelis–Menten approximation. First, a sample data set for (7.2) is generated
with the program fit3.m. Then, the initial velocities from these sample data are
128 8 Fitting the Data

calculated and fitted to a simplified steady state model with the program fit4.m.
Both covariance and correlation matrix reveal that the fitted parameters are not
reliable. The model is further simplified and calculated with fit4r.m. The resulting
parameters show a reasonable standard deviation, but the simplified model is
difficult to relate to the molecular mechanism.
A second fitting strategy fit5.m does not change the model but tries to minimize
the number of parameters with reasonable assumptions. It has to make use of
all experimental information available, so that global fits to progress curves are
performed. The statistical information of the fitting process is analyzed from the
confidence interval and the correlation matrix. Additional information is introduced
to fix parameters which have a high statistical error and high correlation coefficients.
This leads to a reasonable set of parameters, which do not only depend on the experi-
mental data, but also on the additional information invested in the fitting process.

ð7:2Þ

8.4.1 Generating Sample Data (fit3.m)

Sample data are generated in fit3.m by calculating a set of theoretical data and
adding random noise. The substrate concentration is computed from a series of
dilutions by a factor 2 (multiply 0.5) from the initial concentration S00, just
as shown in line 45. The differential equations of scheme (7.2) are solved with
enz4F.m (Sect. 7.4) and called from lsode in line 47.
8.4 Fitting Substrate Inhibition 129

The product concentration P for all times (:) and each substrate concentration
(k) is extracted from the matrix M in line 48. Less than 3% random noise is added
to the product concentration in order to simulate experimental data (Da) in line 52.
The vector t ¼ time had been defined with linspace in line 16 as a row vector.
It is transposed in line 51 to a column vector ct. This agrees with the matrixes P
and Da, where the rows correspond to time and the columns to the substrate
concentration. Note the serial dilution (line 45), so that the first substrate concen-
tration is the largest one. The save command in lines 53 and 54 does not save all
parameters, but selects the vectors time (ct) and substrate concentrations (VS0)
together with the matrices P and Da. The generated data (+) are random variations
of the theoretical values (
) as shown in Fig. 8.4.

8.4.2 Calculating Steady State Equilibria (fit4.m)

Enzyme kinetics can be calculated from initial velocities or progress curves at

different substrate concentrations. Let us begin with initial velocities. The data are
loaded from the file generated in fit3.m. Unlike store, the load command in
line 33 does not require variable names, even when specific variables had been
stored (line 54 in fit3.m). The initial velocities are calculated from the slope of
initial data points in lines 33–35. ct, the column vector for time, and Da, the
corresponding vector of simulated data, had been retrieved with the load com-
mand in line 32.

The logarithms in line 36 are calculated because equilibrium dissociation

constants and rate constants will be varied in a logarithmic scale (pin in line
38). The routine leasqr (lines 40 and 41) calls the function fit4fit which in turn
calculates binding equilibria from reaction scheme (8.5) with the function
Enz4EQF.m.
130 8 Fitting the Data

Sample Progress curves calculated for substrate inhibition (fit3.m)

30

25

20
Concentration of P

15

10

0
0 20000 40000 60000 80000 100000
Time (sec)

Fig. 8.4 Sample progress curves calculated for substrate inhibition. (
) theoretical progress curve
for product concentration P1 in scheme (7.2) with KD1 ¼ 100 mM, KD2 ¼ KD3 ¼ 100 mM.
k1 ¼ k4 ¼ 0.01 mM
1 s
1. k2 ¼ 1 s
1, k3 ¼ 0.05 s
1, k5 ¼ 0.001 s
1. The enzyme concentration
is 10 nM. The substrate concentration varied in a logarithmic scale from 1 mM to 1 mM.
(+) simulated experimental data as variations of P1

ð8:5Þ

A steady state equilibrium is established if the reversible reactions in (8.5) are

much faster than the dissociation of products P1 and P2. Reaction scheme (8.5) is a
simplification of scheme (7.2). It is part of this exercise for simplification strategies.
The equations describing the steady state equilibrium are contained in lines 5–8 of
the function Enz4EQF.m.
8.4 Fitting Substrate Inhibition 131

The fitting procedure itself is based on the function leasqr (lines 40 and 41 of
the main program fit4.m). It requires the function fit4fit.m to calculate the
theoretical initial velocities. Initial velocities are equal to kcat·[ES]. The
substrate–enzyme complexes [ES] are calculated from the steady state equilibria
at different substrate concentrations. Therefore, fit4fit.m solves the set of
equations Enz4EQF in line 17, calculates VES, the vector of [ES] in line 20,
and the initial velocities in line 22. The sum of squares (sos) is calculated
independent of the fitting algorithm in lines 42 and43 of the main program
fit4.m.

Any curve fitting procedure basically is a “try and error” method. The initial
parameters (pin) are used to calculate an initial theoretical curve, from which the
sum of squares (8.1) is calculated. Then the parameters are varied, the next sum of
squares is obtained and compared with the first one, and so forth. Sometimes this
variation of parameters leads to “local minima”, which are combinations of
parameters which do not give the optimal fit. Sometimes this leads to unreasonable
values. Local minima can be avoided by changing pin, the vector of initial
parameters. Unreasonable values can be avoided by setting limits to the fitted
parameters. The MATLAB procedure lsqcurvefit allows selection of upper
and lower limits. The Octave function leasqr does not have this extra, so that it
is included in lines 3–8 of fit4fit.m.
MIN and MAX are the vectors of lower and upper limits of pin. They are defined
in the main program and are transferred to fit4fit.m via the global statement in
line 2. Since pin itself contains the logarithms of parameters, they have to be re-
calculated with the exponent function exp in line 3, before they can be compared to
132 8 Fitting the Data

the limits. Comparison between values is done with the if statement. This state-
ment asks whether a condition is fulfilled. If the condition is fulfilled, a command or
a series of commands is executed. If the condition is not fulfilled, a second if
statement (elseif) may ask if a second condition is fulfilled and allows condi-
tional execution of a second command or series of commands, until the end
statement is encountered. The if statement of fit4fit.m in lines 5–7 has the
following structure:
if (condition1) commands1; elseif (condition2) commands2; end;
Condition 1, ep(I) < MIN(I) is true when the exponent of the fitted loga-
rithm of a parameter is smaller than MIN. In this case, the fitted parameter will be
set equal to the logarithm of MIN with the command pin(I) ¼ log(MIN(I)).
Condition 2 is true when the exponent of the fitted parameter is larger than MAX, and
consequently it then will be forced to be equal to log(MAX). For the actual
calculations, the varied parameters pin have to be translated into meaningful
rate or equilibrium constants. This is done within fit4fit.m in lines 9–13. Calcu-
lating the exponent from the logarithm has the additional advantage that the
constants never reach negative values. Negative concentrations or rate constants
simply do not make sense.
Note that the function leasqr in lines 40 and41 of the main program fit4.m
expects column vectors as arguments. The transpose operator ’ in line 22 of
fit4fit.m is used for this purpose. Also note that the resulting parameters from
pin in lines 9–12 also are defined in global. This may seem unnecessary, but
sometimes one likes to repeat the fit and keep one parameter fixed. This can easily
be achieved without re-writing the whole program, simply by inserting the percent
symbol % in front of a line. This turns the line into a comment. For example, line 9
written as %KD1 ¼ exp(pin(1)); will not be executed. Because KD1 is
included in the global statement, this variable can be used inside the function
fit4fit.m. Its value then is not varied and kept at the initial value defined in line 24
of fit4.m. The results of fit4.m are shown in Fig. 8.5.
For MATLAB users, the programs fit4M.m and fit4fitM.m are slightly differ-
ent. The MATLAB routine lsqcurvefit allows the selection of MIN (lb) and
MAX(ub) values directly, so that the if and elseif statements in lines 3–8 are
not required. The function lsqcurvefit is rather flexible, but this flexibility
requires options to be set. This is done in line 42 of the program fit4M.m with
the function optimset.

In general, MATLAB users may have to spend some time trying to find the best
options for the function lsqcurvefit. This is similar to finding the optimal solver
for differential equations. GNU Octave usually runs quite smoothly with its default
parameters both for leasqr and for lsode.
8.4 Fitting Substrate Inhibition 133

Steady state model for substrate inhibition (fit4.m)

0.0003

0.00025

0.0002
Initial velocity

0.00015

Parameters obtained from fitting

0.0001
KD1 = 661.4 µM (± Inf)
KD2 = 0.0205 (± Inf)
KD4 = 93.7 µM (± 4.4726e+089)
5e-005 kcat = 2.6 (± Inf)
Sum of Squares 2.0558e-010

0
0 0.5 1 1.5 2 2.5 3
log(Substrate concentration (µM))

Fig. 8.5 Multi-parameter fit of initial velocities fitted to a steady state approximation of substrate
inhibition (8.5). The “experimental” data from fit3.m [scheme (7.2)] were fitted to the steady
state approximation with the program fit4.m. The fitted values are shown in the plot. The
calculated standard deviations are excessively high, so that the resulting parameters are not
reliable

Figure 8.5 shows a good fit of theoretical curves (straight line) to “experimental”
initial velocities (o). However, the standard deviations are excessively high, so that
the resulting parameters are not reliable. The standard deviations of the fitted
parameters are calculated in line 44 of the main program fit4.m from the diagonal
elements of the covariance matrix covp (result from leasqr in lines 40 and41).
These standard deviations are the standard deviations of the logarithmic scale
used for the fitting process. They are translated to delta values of the linear
scale in line 45.
The Octave function leasqr also returns a correlation matrix corp (lines 40
and 41). The matrix is displayed by typing corp in the Octave terminal window
after running the program. A correlation matrix contains as many rows and columns
as there are parameters. The respective correlation coefficients are shown at the
respective intersections of rows and columns. The correlation of parameter 2 (KD2)
and 4 (kcat), for example, is the matrix element corp(2,4), which is identical to
the matrix element c orp(4,2). Correlation coefficients can vary between
1
and 1. The absolute value gives the amount of correlation ranging from 0 (not
correlated) to 1 (strongly correlated). Positive correlation coefficients of two
parameters indicate that an increase of parameter 1 can be compensated with an
increase of parameter 2, whereas negative values indicate that an increase of
134 8 Fitting the Data

Table 8.1 Matrix of KD1 KD2 KD4 kcat

correlation coefficients from
the fit of Fig. 8.5 calculated KD1 1
1 1 1
with the program fit4.m KD2
1 1
1
1

KD4 1
1 1 1

kcat 1
1 1 1

parameter 1 can be compensated by a decrease of parameter 2. Each parameter is

correlated with itself, so that the diagonal elements of any correlation matrix are 1.
Every time a correlation coefficient with an absolute value near 1 appears outside
the diagonal, the values are heavily correlated and cannot be regarded as indepen-
dent results. The correlation matrix resulting from fit4.m is shown in Table 8.1.
Indeed, the correlation matrix in Table 8.1 shows heavy correlation of
parameters, which explains why the standard deviations of the parameters are so
high. Here, we follow the strategy of model simplification. We therefore will not try
to keep one parameter fixed, but instead will further simplify the model. Reaction
scheme (8.6) is the simplest conceivable model to calculate substrate inhibition
under steady state conditions.

ð8:6Þ
The simplified scheme (8.6) is calculated with the program fit4r.m, which calls
the fitting routine fit4rfit.m, which in turn solves the set of equations
Enz4EQRF.m. The result is shown in Fig. 8.6. These programs are very similar
to the originals (fit4.m, fit4fit.m and Enz4EQF.m) and need not be discussed.
8.4 Fitting Substrate Inhibition 135

Simplified steady state model for substrate inhibition (fit4r.m)

0.0003

0.00025

0.0002
Initial velocity

0.00015

0.0001 Parameters obtained from fitting

KD1 = 13.3 µM (± 0.9)
KD4 = 95.5 µM (± 6.7)
5e-005 kcat = 0.053 (± 0.002)
Sum of Squares 2.0558e-010

0
0 0.5 1 1.5 2 2.5 3
log(Substrate concentration (µM))

Fig. 8.6 Multi-parameter fit of initial velocities for substrate inhibition. The “Experimental” data
from fit3.m were fitted to the steady state approximation (8.6) by means of fit4r.m. The fitted
parameters and their standard deviations are printed within the plot. The correlation coefficients
are listed in Table 8.2

Table 8.2 matrix of correlation coefficients from the fit of Fig. 8.6 calculated with the program
fit4r.m
KD1 KD4 kcat
KD1 1.0
0.85 0.94
KD4
0.85 1.0
0.94
kcat 0.94
0.94 1.0

The correlation matrix corp for the fit of Fig. 8.6 is shown in Table 8.2. The fit
is excellent, and gives the same sum of squares as for reaction scheme (8.5). The
errors are smaller, so that individual parameters can be determined with much better
precision. The parameters in Table 8.2 still are correlated, but not as heavy as in
Table 8.1. Reaction scheme (8.6) cannot be simplified any more.
The numeric results of a fit to reaction scheme (8.6) could not have been inferred
from the calculated reaction of scheme (7.3). Or to summarize it more ruthlessly:
The simplified reaction scheme (8.6) may be appropriate for fitting initial velocities
with a minimum of parameters, but does not reflect the underlying mechanism.
136 8 Fitting the Data

8.4.3 Global Fit of Progress Curves (fit5.m)

A global fit of progress curves is the most rigorous method to analyze enzyme
kinetics. It takes into account all available experimental information. For the global
fit calculated with fit5.m, these are the progress curves of the matrix Da generated
by fit3.m and stored in SubInhib.mat. The octave function leasqr allows
entering sets of different experiments only when they are arranged into one single
column vector. The matrix (spreadsheet) Da (Fig. 8.4) is arranged in N2 ¼ 12 rows,
corresponding to the 12 concentrations, and N ¼ 30 columns, corresponding to 30
time points. This matrix is turned into a single column vector of 12  30 ¼ 360
rows with the commands in lines 46–50. The name of the linearized data set is
lDa. The function leasqr requires a vector of the same size for the independent
variable (the time). Such a vector is created in line 46 with the zeros command. It
creates a column vector lt of N*N2 rows containing zeros. This vector lt is just a
dummy.

The real time vector is ct, which is handed over to the function fit5fit by means
of the global statement in lines 23–24 of fit5.m, just as lines 2–3 of fit5fit.m,
listed below. Of course, the theoretical fit has to have the same structure as the
experimental data in the main program, and therefore the fitted values mes returned
from the function fit5fit consist of same single column vector format, lines
27–30. mes is calculated from the matrix of product concentrations P(:,k) at
different substrate concentrations k (line 29).
Reaction scheme (7.2) uses nine different rate constants, but the experiments do
not allow the determination of all of them. Enzyme kinetics is measured under
steady state conditions, so that the association rate constants k1 and k4 cannot be
fitted. They were set arbitrarily to the relatively high value of 0.1 mM
1 s
1.
Likewise, enzyme kinetics with excess of substrate do not allow the measurement
of the back reaction, so that the equilibrium dissociation constants KD2 and KD3
could not be fitted and were set arbitrarily to a high value (10 mM). This leaves five
parameters to be fitted, namely KD1, KD4, k2, k3 and k4. The logarithms of their
initial estimates are the elements of the vector pin, and the exponent function must
be used to extract the parameters from pin in lines 10–15 of fit5fit.m. The rate
constants for the back reactions are calculated from the association rate constants
and the equilibrium dissociation constants in lines 17–21, so that they can be
passed to the differential equations (enz4F.m) with the global statement.
8.4 Fitting Substrate Inhibition 137

One important feature is missing in our example of theoretically calculated data:

There are no real experimental errors. Typically, one would expect background
variations between the different kinetic experiments. They have to be included as
additional parameters. Since they only concern one subset of the experimental data,
these fluctuations are not correlated with the other parameters in the fitting proce-
dure. It is better to consider the variation between experiments as additional fitting
parameters than to ignore it and get meaningless results.
Once the data have been computed in the global fitting routine as a linear vector
fcurve (line 51), they have to be re-arranged in lines 59–62, so that theo(j,k)
is the theoretical value for each data point Da(j,k), to be used in line 65 for the
plot command.
138 8 Fitting the Data

The statistical information of the covariance matrix covp is used to calculate the
95% confidence interval for the fitted parameters. This is done in lines 56–58.
Remember, that for a Gauss distribution 95% of the events are expected in an area
of  two times the standard deviation around the mean value. In lines 57 and 58,
low and high therefore denote the lower and upper limits of the confidence
interval. The matrix of correlation coefficients corp, computed with leasqr, is
saved in ascii format in line 53 of fit5.m. It can be retrieved with any editor, or
one simply can type corp in the Octave window. The matrix only contains the
numbers of Table 8.3, not the names of the parameters. The parameter names
correspond to the order of parameters in the vector pin (line 43).
The number of data displayed in Fig. 8.7 becomes rather large, so that their
length is reduced by means of formatted output. This is specified for the num2str
commands in lines 72–81. For example, the command num2str(KD1,'%
4.1f') used in line 72 turns the number KD1 into a string. The output format
'%4.1f' uses four parameters, the percent sign (%) to specify that a conversion to
formatted output is expected, a letter (f, in this case) to specify conversion to a
fixed point notation, a number (4) to specify the minimum field width and a number
(1) to specify the decimal precision. This gives the resulting string 51.9 shown in
Fig. 8.7. Other useful output conversions are %e or %E to specify exponential

Table 8.3 Matrix of correlation coefficients from the fit of Fig. 8.7 calculated with the program
fit5.m
KD1 KD4 k2 k3 k5
KD1 1.0 1.0 1.0
1.0
0.02
KD4 1.0 1.0 1.0
1.0
0.04
k2 1.0 1.0 1.0
1.0
0.02
k3
1.0
1.0
1.0 1.0 0.03
k5
0.02
0.04
0.02 0.03 1.0
8.4 Fitting Substrate Inhibition 139

Global Fit of Substrate Inhibition(fit5.m)

30
Parameters obtained from fitting
KD1 = 33.6 µM // 1.4 - 792.1
25 KD4 = 77.8 µM // 25.2 - 240.5
k2 = 0.178 s–1 // 0.008 - 4.23
k3 = 0.065 s–1 // 0.021 - 0.20
k5 = 0.0010 s–1 // 0.0009 - 0.0011
20
Concentration of P

Sum of Squares 2.4305

15

10

0
0 20000 40000 60000 80000 100000
Time (sec)

Fig. 8.7 Global fit of substrate inhibition. The “Experimental” data from fit3.m (KD1 ¼ KD4
¼ 100 mM, KD2 ¼ KD3 ¼ 100 mM. k1 ¼ k4 ¼ 0.01 mM
1 s
1. k2 ¼ 1 s
1, k3 ¼ 0.05 s
1,
k5 ¼ 0.001 s
1) were fitted to reaction scheme (7.2). (
) calculated progress curve (+) simulated
experimental data from fit3.m. The resulting parameters are shown inside the plot together with a
95% confidence interval. The correlation coefficients are listed in Table 8.3

notation and %g or %G for either normal (fixed point) or exponential notation,

depending on the magnitude.
Running the program fit5.m may take several minutes, depending on the
computer. Figure 8.7 shows a perfect global fit of the progress curves of substrate
inhibition. The confidence intervals, however, were rather large. Note that “experi-
mental” data were generated with random noise in fit3.m, so that each data set is
different. Only k5 could be determined with a reasonable accuracy. Likewise, the
only rate constant which is not correlated to the other ones is k5. Indeed it had been
reported [8] that this step has a profound influence on the characteristics of substrate
inhibition.
Rather than simplifying reaction schemes as in Sect. 8.4.2, we will use a
different strategy. We will stick to scheme (7.2), but decrease the number of
parameters. At this stage, one typically would take plausible estimates from the
literature for individual rate constants or affinities. Here we take the linear regres-
sion of the initial part of the Lineweaver–Burk plot (enz4.m, Fig. 7.9) and assume
that our KD1 corresponds to the KM estimated there. The global fit fit5b.m therefore
keeps KD1 fixed to 10.8 mM. The result is shown in Fig. 8.8.
The quality of the fit shown in Fig. 8.8 is excellent. The 95% confidence intervals
of the parameters are rather low, so that one may tend to believe in the results.
140 8 Fitting the Data

Global Fit of Substrate Inhibition(fit5b.m)

30

Parameters obtained from fitting

KD4 = 18.2 µM // 16.4 - 20.1
25
k2 = 0.057 s–1 // 0.057 - 0.058
k3 = 0.28 s–1 // 0.25 - 0.31
k5 = 0.0010 s–1 // 0.0009 - 0.0011
20 Sum of Squares 2.4278
Concentration of P

15

10

0
0 20000 40000 60000 80000 100000
Time (sec)

Fig. 8.8 Global fit of substrate inhibition. The “Experimental” data from fit3.m (KD1 ¼ KD4
¼ 100 mM, KD2 ¼ KD3 ¼ 100 mM. k1 ¼ k4 ¼ 0.01 mM
1 s
1. k2 ¼ 1 s
1, k3 ¼ 0.05 s
1,
k5 ¼ 0.001 s
1) were fitted to reaction scheme (7.2) with KD1 ¼ 10.8 mM taken from the
approximation of Fig. 7.9. (
) calculated progress curve (+) simulated experimental data from
fit3.m. The resulting parameters are shown inside the plot together with a 95% confidence
interval. The correlation coefficients are listed in

However, the fitted values do not correspond to the original ones. The “experimen-
tal” data were generated with KD4 ¼ 100 mM, but were fitted with 18.2 mM; k2 was
fitted with 0.057 s
1 instead of the expected 1 s
1; k3 was fitted to 0.28 s
1 instead
of 0.05 s
1. Only k5 was determined correctly as 0.001 s
1.
The reason for this discrepancy lies, of course, in the assumption that KD1
corresponds to the Michaelis constant KM of 10.8 mM determined in enz4.m. It is
obvious from the intrinsic parameters listed in Fig. 7.9 that this assumption is wrong.
Since all these parameters are correlated, as shown in Table 8.3, one wrong number
leads to a domino effect, giving wrong values for all correlated parameters. Only k5 is
not correlated and therefore was not affected by the wrong choice of KD1.
The final results in all sample programs are plotted with the plot command,
which is based on gnuplot, a powerful tool on its own [9]. For scientists who are not
familiar with it may be easier to save their results in ascii format (see, for example,
lines 39 and 40 of EQ1.m) and produce publication quality plots with a known
software (Table 8.4).
Global fits are the most rigorous type of quantitative data analysis in life sciences.
A successful fit always proves that the underlying molecular model is sufficient to
explain all fitted experimental observations. The parameters from a multi-parameter
References 141

Table 8.4 Matrix of correlation coefficients from the fit of Fig. 8.8 calculated with the program
fit5b.m
KD4 k2 k3 k5
KD4 1.0 0.96
1
0.58
k2 0.96 1.0
0.97
0.43
k3
1
0.97 1.0 0.54
k5
0.58
0.43 0.54 1.0

fit need not be unique. When parameters are correlated, there may be infinite
combinations of parameters which fit the data equally well. Reducing the number of
parameters increases their reliability, but one invalid assumption can corrupt the lot.

References

1. http://en.wikipedia.org/wiki/Simple_linear_regression
2. http://en.wikipedia.org/wiki/Correlation_and_dependence
3. Croxton FE, Cowden DJ, Klein S (1968) Applied general statistics. Pitman, London, p 625
4. Dietrich CF (1991) Uncertainty, calibration, and probability: the statistics of scientific and
industrial measurement. Inst of Physics Pub, London, p 331
5. Aitken AC (1936) Statistical mathematics. Oliver and Boyd, Edinburgh, p 95
6. Rodgers JL, Nicewander WA (1988) Thirteen ways to look at the correlation coefficient. Am
Stat 42:59–66
7. Hofer P, Fringeli UP (1981) Acetylcholinesterase kinetics. Biophys Struct Mech 8:45–59
8. Krupka RM, Laidler KJ (1961) Molecular mechanisms for hydrolytic enzyme action. J Am
Chem Soc 83:1445–1460
9. Janert PK (2009) Gnuplot in action. Understanding data with graphs. Manning Publications Co,
Greenwich, CT
.
Chapter 9
Appendix: Installing GNU Octave
for Mac OS and Linux

9.1 Installation Instructions for Mac OS X

Octave for Mac OS X can be downloaded from the GNU webpage http://www.gnu.
org/software/octave/download.html. Select Octave.app for Mac OS X and with
regard to your processor choose the ocatve-3.2.X-i386.dmg (for Macs newer than
2006 based on Intel processors) or octave-3.2.X-ppc.dmg (for older Macs before
2006 based on Power PC processors). Open the downloaded dmg-package with a
double click and drag the icon “Octave” from the open window to the icon
“Applications”. That will install Octave. The sample programs in this book require
Gnuplot and additional packages.
Gnuplot is installed from the folder “Extras” in the package octave.app.dmg,
which had been opened before, by double clicking the gnuplot package. This will
open the gnuplot installation package. There, the icon “Gnuplot” has to be dragged
to the icon “Applications”.
Ocatave can be expanded with different additional packages. These require the
“XCode Apple’s developers tools” for installing. These Apple tools can be installed
from the Mac OS X installation DVD listed under “optional installs” or downloaded
from Apple’s web site after registration. One has to double click the package
“Xcode.mpkg” and follow the installation instructions using the default values.
Afterwards, the “X11” package from the “Optional Installs.mpkg” in the same
folder has to be installed. Double click the icon and follow the installation
instructions, choose “Custom Install”, select the package “X11” found under
“Applications” and continue the installation, just follow the instructions.
After installing “XCode” and “X11” your system is ready to install the required
Octave packages “miscellaneous”, “io”, “struct" and "optim”. They are found on
the octave website (http://octave.sourceforge.net/packages.php). Mac OS X saves
the files in the folder “Downloads” in your home directory. Having done this you
can run octave now from the applications. A white window, and the terminal
console will open. Now type in

H. Prinz, Numerical Methods for the Life Scientist, 143

DOI 10.1007/978-3-642-20820-1_9, # Springer-Verlag Berlin Heidelberg 2011
144 9 Appendix: Installing GNU Octave for Mac OS and Linux

The package in this example is called “miscellaneous” in version 1.0.10 (latest

version in time of print). If you are downloading a newer version change the
command line according to your package version. Install the other packages
according to your download path and their version names:

You can ignore warnings but be careful for error messages. There should be no
error messages if you have installed all required packages and followed the above
procedure. Verify that all packages are installed by typing pkg list, and the
output should be similar to

Congratulations! Octave can now be used with Mac OS for all sample programs
of this book.

9.2 Installation Instructions for Linux

The operating system GNU/Linux comes in many distributions. Here, the installa-
tion for “Ubuntu Linux” and all other distributions based on the debian package
archive is shown. Simply start your Linux terminal and run

This will install all necessary packages for running all examples in this book.
After installing you will find Octave in the Gnome menu. Go to “Applications”,
then “Programming” and click on “GNU Octave”. If you are using a terminal you
just have to type “octave”. Now you can test if all necessary packages have been
installed by typing “pkg list” in the octave window. The output shows all
additional packages for octave with their version numbers similar to the following:
9.2 Installation Instructions for Linux 145
.
Index

A correlation coefficient, 122

acetylcholine receptor, 90 correlation matrix, 120, 133, 134
acetylcholinesterase, 107 covariance matrix, 133, 138
addpath, 40 covp, 133
allosteric, 57 Ctrl‐C, 40
allosteric activator, 62
allosteric effector, 63, 64
allosteric, multiple, 67 D
analytical solutions, 15 data input, 125
arrays, 38 data output, 125
ascii, 48, 138 data, generation of samples, 128
axis, 47 deacetylation, 107
debian, 144
decimal precision, 138
B difference quotient, 73
back box, 1 differential equations, 10, 74
binding kinetics, 71, 73 dilution experiment, 81
dilution series, 124
djpeg, 48
C dlmread, 125, 126
chase experiment,81, 91 dmeta, 48
clear, 45 dose activity curve, 110, 117
colon, 74, 110 dose‐response curve, 23, 64, 70
colors, 52, 98 double reciprocal plots, 17
colors, fill, 99
column vector, 38
comment, 44 E
competitive, 114, 115 Eadie‐Hofstee plot, 20
concentration of bound ligand, 51 element by element division, 47
concentration of complexes, 51 element by element multiplication,
condition, 132 35, 39
confidence interval, 138 elseif, 131, 132
conformational state, 58 emf, 48
continuation marker, 47, 87 end statement, 47
corp, 120, 133, 138 endfor, 47
correlated parameters, 123 endif, 47
correlation, 119 enzyme kinetics, 22, 97

H. Prinz, Numerical Methods for the Life Scientist, 147

DOI 10.1007/978-3-642-20820-1, # Springer-Verlag Berlin Heidelberg 2011
148 Index

enzyme kinetics, inhibition, 111 kinetics, ordered binding, 86

equilibrium, 7, 15, 43
equilibrium two sites, 50
equilibrium, coupled, 8, 54 L
equilibrium, intrinsic constants, 10 lag‐phase, 79
exit, 40 leasqr, 120, 121
exponential decay, 25 least squares, 119
legend, 48
limit, 131
F line, 106
facilitated dissociation, 84 Lineweaver‐Burk plot, 21, 22, 106, 110, 111
filename, 45 linreg, 105, 110, 121
fitting, 119, 127 linspace, 35, 45
fixed point notation, 138 Linux, 144
for statement, 46 literature, 40
formatted output, 138 load, 126, 129
fsolve, 43, 44, 46 logarithmic scale, 129
function, 44 logistic function, 24, 94
function linreg, 105 logspace, 103
lsode, 71, 74, 98
lsqcurvefit, 119, 120
G
global, 44
global fit, 124, 125, 136 M
gnuplot, 140, 143 Mac OS X, 143
H MATLAB, 40, 72, 132
half life, 25 matrices, 38
half‐logarithmic plot, 26 matrix to vector, 135
Hill coefficient, 23, 70, 95, 117 max(), 57, 105, 106
Michaelis constant, 21
Michaelis‐Menten, 21, 97, 101, 102
I minimum field width, 138
if statement, 132 mkdir, 40
implicit conversion, 47 Multi‐parameter fits, 119
inflection point, 117 multiple exponential fit, 28
inhibition, reversible, 111 MWC model, 57, 60
inhibitor, 52
initial velocity, 98, 102, 109, 131
initial velocity, association kinetics, 27 N
installation, 31 nested loop, 55
irreversible inhibition, 91 nlinfit, 31
noncompetitive, 114
Notepadþþ, 32
J num2str, 100
jpg, 48 numerical artifact, 117
numerical solutions, 12, 50

K
kinetics, association, 75 O
kinetics, competitive binding, 81 Octave, 31, 40
kinetics, conformational change, 77 Octave version 3.2.4, 47, 48
kinetics, dissociation, 74 ode23s, 71, 98
kinetics, obstructed dissociation, 87 ode45, 71, 72, 98
Index 149

optimset, 132 Statistics Toolbox, 31

outlier, 117 steady state, 109
overlapping sites, 86 stiff differential equations, 90, 98
store, 129
string, 39
P substrate inhibition, 107, 127
parameters, set of correlated, 124 sum, 105
path, 40 sum of squares, 131
pin, 120
pkg install, 144
plausible scheme, 98 T
plot, 34, 47, 98 ternary complex, 57
plot style, 52 text, 40, 100
print, 48 title, 40, 46
product formation, 103 transient binding patches, 70
product inhibition, 107 transpose operator, 38, 116
progress curves, 97 tutorial, 39, 40
property value pair, 100
pseudo first order, 28
U
R Ubuntu, 144
rand, 121, 128 uncompetitive, 116
random noise, 128 units, normalized, 100
random numbers, 121
range of values, 110
reaction, first order, 5 V
reaction, second order, 5 variable, 45
regression, nonlinear, 120 variable temp, 66
regression, simple linear, 105 vectors, 38
row vector, 38
W
S warning, 13
sample programs, 2, 3 weak substrate, 99
save, 48, 126, 128
scalar product, 38, 39
Scatchard plot, 19, 61 X
screening, 92 xlabel, 39, 47
sequential, 8–10, 68, 87, 90 xlsread, 126
simple allosteric model, 67
simplify a model, 134
size, 105 Y
solver, 71 ylabel, 47
standard deviations, 133