13/10/2015

Computational Modelling in Physics,

Chemistry and Biochemistry

Molecular Modelling

Unit 1. Multi-scale Description of Atomic and Molecular

Systems

J. Torras , Univ. Politècnica Catalunya (2015)

Contents
 1.1 System Description at the Quantum and Classical level
 1.2 Foundations of statistical mechanics and molecular
modeling
 1.3 Calculation of the System Energy as a Common Starting
Point
 1.3.1 Quantum Methods: Modelling Subatomic Particles
 1.3.2 Classical Methods: Modelling Classical Particles
 1.3.3 Numerical Methods: Continuum Modelling
 1.4 Hybrid QM/MM methods
 1.5 The Simulation Concept. Simulation Types:
 1.5.1 Stochastic: Monte Carlo Method
 1.5.2 Mechanical: molecular dynamics, Brownian dynamics

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

1
 13/10/2015

1.1 System Description at the Quantum

and Classical level
 Computational Quantum Chemistry
 Numerical Solution to Schrodinger equation
 Computationally intensive for small molecules
 In principle, yield exact electronic structure and energy as limiting
case of increasingly accurate method (HF, MP2, MP4, ….)
 DFT (Density Function Theory) is approximate but fast

 Molecular Simulation
 Monte Carlo Simulation
 Molecular Dynamic Simulation
 Both require intermolecular and intramolecular potentials
(force field) as input

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

Molecular Modelling
 Theoretical Foundations
 Newton (1643-1727)  Classical equations of motion F(t)=ma(t)
 Schrödinger (1887-1961)  Quantum mechanical equations of motion
 Boltzmann(1844-1906)  Foundations of statistical mechanics

 Simulation milestones
 Metropolis(1953)  First Monte Carlo (MC) simulation of a liquid (hard spheres)
 Wood(1957)  First MC simulation with Lennard-Jones potential
 Alder (1957)  First Molecular Dynamics (MD) simulation of a
liquid (hard spheres)
 Rahman (1964)  First MD simulation with Lennard-Jones potential
 Lifson & Warshel (1968)  First consistent Force Field (Alkanes)
 Levitt & Lifson (1969)  Molecular Mechanics(MM) First Protein structure refinement
 Warshel & Levitt (1976)  QM/MM models for catalysis
 McCammon, Gelin &
 Karplus (1977)  First MD simulation of proteins
 Karplus(1983)  CHARMM general purpose FF and MD program
 Kollman (1984)  AMBER general purpose FF and MD program
 Car-Parrinello (1985)  First MD simulation using full QM calculations

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

2
 13/10/2015

1.3 Calculation of the System Energy as a

Common Starting Point
 Modelling by balls & sticks
 You previously have to know its molecular
structure
 No temperature
 No Normal modes of vibration
 No Internal rotations
 It gives us some clues of which Proline amino acid model
conformations are not allowed.
 Doesn’t help us to identify the “true”
molecular geometry

 System Energy
 Is the key to know the structure stability
and the final molecular geometry
3D viewer of molecular structures

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

1.3 Calculation of the System Energy as a

Common Starting Point
 Experimental methodology of molecular structure
resolution
 Gas-phase molecules
 microwave spectroscopy and electron diffraction
 Solid state
 X-ray and neutron diffraction
 Proteins
 X-ray crystallography
 limited by the difficulty of getting some
proteins to form crystals
 Nuclear magnetic Resonance (NMR)
 Can only be applied to relatively small
protein molecules
 Additional software on structural modelling resolution
 Protein recognition based in previous knowledge.

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

3
 13/10/2015

1.3 Calculation of the System Energy as a

Common Starting Point
 Steps in protein modelling
 Similarity search
 Detection of suitable templates sequences.
 BLAST algorithm (Basic Local Alignment Search
Protein)
 J. Molecular Biology 1990; 215(3):403
 Comparative modelling (secondary structure)
 A sequence of amino acids is compared with a very
similar template protein of known structure.
 Annu. Rep. Med. Chem 2004; 39 :39020
 Fold recognition (Protein threading)
 Model proteins with the same ”fold” as proteins of
known structures (homologous protein is not
known)
 Protein Sci. 1996 ; 5(5):947-55
 Ab initio modelling (De novo protein structure
prediction)
 Tertiary structure is predicted from its amino acid
sequence.
 Annu. Rev. Biophys. Biomol. Struct 2001; 30:173-89
Baker & Sali, Science 2001; 294:93-96

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

1.3 Calculation of the System Energy as a

Common Starting Point
 Molecular structure database
 Cambridge Structural Database (CSD)
 http://www.ccdc.cam.ac.uk
 750.000 inputs
 Organic and metal-organic
compounds
 Protein Data Bank (PDB)
 http://www.rcsb.org
 110790 biological structures
 3D biological macromolecular structural data

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

4
 13/10/2015

1.3 Calculation of the System Energy as a

Common Starting Point
 To understand the behavior of complex systems need:
 The potential surface on which the atoms move
 The laws of motion for the atoms

ATP Synthase
(natural molecular motor)
uses the power from a
proton gradient to force the
“F1” motor to generate ATP
molecules

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

1.3 Calculation of the System Energy as a

Common Starting Point
 Molecular Simulation
 interrelationship between theory and experimental data
 All chemical behavior is mainly located on the force field
 Crucial to describe big system at low level simulation

Ab initio
calculations
Force Fields Methods
Experimental
data

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

5
 13/10/2015

1.3 Calculation of the System Energy as a

Common Starting Point
 Theory and Experimental observables

Quantum Mechanics Approximate Theory

Statistical Physical
Mechanics Properties
Classical Mechanics Molecular Simulation

O T,P, U,H,G,S, m, Cp, …

O
O Pressure Plot

28

26

N 24

Pressure
22

20

18

16

14
0 200 400 600 800 1000 1200

Time

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

1.3 Calculation of the System Energy as a

Common Starting Point
 Advances in Hardware
and Software
 Moore’s Law
 Computing speed double
 every 18 month
 Order of magnitude every 5
years
 Add 2-3 orders of
magnitude using
parallelization

 Theory
 Theory and methodology
improvement

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

6
 13/10/2015

1.3 Calculation of the System Energy as a

Common Starting Point
 Computer simulation and Experiments
“perfect tandem”
 Advances in Hardware and Software

 Experimental Cost
 Labor intensive,
 High capital cost

 Theory
 Labor intensive
 Do graduate students and lab equipment
improve by an order of magnitude every five
years ?
Johnson Nat. Chem. 2009; 1:8-9
 Synergy among Theory and Experiment
 Nature Chemistry 2009; 1:8-9
http://www.nature.com/nchem/journal/v1/n1/full/
nchem.145.html
J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

1.3 Calculation of the System Energy as a

Common Starting Point
 Time scale and space scale in molecular simulations
Time

years Engineering Whole Process

Design
hours
FEM, Process
minuts Unit Operations
Simulation
seconds Mesoscale
Modeling Segments
microseconds
Molecular Atoms, moleculaes
nanoseconds Mechanics
picoseconds
Quantum electrons
femtoseconds Mechanics
Distance
1A 1nm 1m 1mm meters

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

7
 13/10/2015

1.3.1 Quantum Methods: Modelling

Subatomic Particles
 Quantum Scale
 Solve Schrodinger Equation
 Ψ Ψ
 Electronic density
 Electronic states
 System energy

 Electronic structures of atoms and

molecules
 Geometry optimization Highest Occupied Molecular Orbital (HOMO )
of an aniline oligomer
 Interaction energies

 Results are often used in the design

of force field providing connection
to next scale
 Modelling of interactions Non Covalent Interactions (NCI) between
3,4-ethylenedioxythiophene (EDOT) and proline amino acid

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

1.3.1 Quantum Methods: Modelling

Subatomic Particles
 File Formats
 Cartesian Coordinates
 PDB file
 Z-matrix
 Utilities
 File Format converter
 Open Babel
 http://www.openbabel.org
 3D visualizers
 VMD
 Pymol VMD visualizer window showing the tryptophan 7-
halogenase enzyme, highlighting the active site of
 Jmol enzyme.
 Accelrys

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

8
 13/10/2015

1.3.1 Quantum Methods: Modelling

Subatomic Particles
 Cartesian Coordinates
 The easiest and most extended format.

 PDB file
 Generalized for big and complex molecules such as
Proteins
 Introduce connectivity among atoms

 Z-matrix
 (Internal Coordinates)
 Great advantage in highly symmetrical systems.

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

1.3.1 Quantum Methods: Modelling

Subatomic Particles
 Cartesian Coordinates H H

 XYZ file format (*.xyz extension)

H H

 File Specification:
 http://openbabel.org/wiki/XYZ_%28for H H
mat%29
12
bencene.xyz Energy: -145706.8891707
C 0.00000 0.00000 1.40029
<number of atoms> C 0.00000 1.21284 0.70025
comment line C 0.00000 1.21284 -0.70025
C 0.00000 0.00000 -1.40029
<element> <X-coord> <Y-coord> <Z-coord>
C -0.00000 -1.21284 -0.70025
... C -0.00000 -1.21284 0.70025
H 0.00000 0.00000 2.48638
H 0.00000 2.15324 1.24296
H 0.00000 2.15324 -1.24296
H 0.00000 0.00000 -2.48638
H -0.00000 -2.15324 -1.24296
H -0.00000 -2.15324 1.24296

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

9
 13/10/2015

1.3.2 Classical Methods: Modelling

Subatomic Particles
 PDB File
 PDB file format (*.pdb extension)
 http://www.wwpdb.org/documentation/file-format-content/format33/v3.3.html
 http://www.cgl.ucsf.edu/chimera/docs/UsersGuide/tutorials/pdbintro.html
 ATOM (A6,I5,1X,A4,A1,A3,1X,A1,I4,A1,3X,3F8.3,2F6.2,10X,A2,A2 )
HEADER CONECT 1 2 6 7
REMARK -- No Title Specified -- CONECT 2 1 3 8
HETATM 1 C 1 0.000 0.000 0.000 C CONECT 3 2 4 9
HETATM 2 C 1 0.000 0.000 1.400 C CONECT 4 3 5 10
HETATM 3 C 1 1.212 0.000 2.100 C CONECT 5 4 6 11
HETATM 4 C 1 2.425 0.000 1.400 C CONECT 6 1 5 12
HETATM 5 C 1 2.425 0.000 0.000 C CONECT 7 1
HETATM 6 C 1 1.212 0.000 -0.700 C CONECT 8 2
HETATM 7 H 1 -0.941 0.000 -0.543 H CONECT 9 3
HETATM 8 H 1 -0.941 0.000 1.943 H CONECT 10 4
HETATM 9 H 1 1.212 0.000 3.186 H CONECT 11 5
HETATM 10 H 1 3.365 0.000 1.943 H CONECT 12 6
HETATM 11 H 1 3.365 0.000 -0.543 H END
HETATM 12 H 1 1.212 0.000 -1.786 H

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

1.3.2 Classical Methods: Modelling

Subatomic Particles
 Z-matrix # B3LYP/6-31G Opt

-- No Title Specified --
 Extensive use in several QM programs
0,1
 Internal Coordinates C
C, 1, R1

 Allows symmetric optimization C, 2, R1, 1, A1

C, 3, R1, 2, A1, 1, D1

 File Specification (Gaussian 09): C, 4, R1, 3, A1,

C, 5, R1, 4, A1,
2, D1
3, D1
H, 1, R2, 6, A2, 5, D2
 http://www.gaussian.com/g_tech/g_ur/c_zm H, 2, R2, 1, A2, 6, D2

at.htm H, 3, R2, 2, A2,

H, 4, R2, 3, A2,
1, D2
2, D2
H, 5, R2, 4, A2, 3, D2
H H
H, 6, R2, 5, A2, 4, D2
Variables:
R1=1.400
R2=1.086
A1=120.00
H H
A2=120.00
D1=0.00
D2=180.00

H H

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

10
 13/10/2015

1.3.2 Classical Methods: Modelling

Classical Particles
 Atomic and Molecular Scale
 Based on a Force Field
 Mainly describe all the chemical behavior

 Usual methodologies
 Molecular Mechanics
 Molecular Dynamics
 Monte Carlo Methods
 Thermophysical properties and transport properties using
statistical mechanics

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

1.3.2 Classical Methods: Modelling

Classical Particles
 Force Field
 Molecular Potential energy among system particles

Michael Levitt, Nobel Lecture (2013)

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

11
 13/10/2015

1.3.2 Classical Methods: Modelling

Classical Particles
 Motion over energy surface.

Michael Levitt, Nobel Lecture (2013)

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

1.3.3 Numerical Methods: Continuum

Modelling
 Finite Elements
 Discretization of initial domain

 Assembly
 Obtain the solution equations for the
system by combining the solution
equations of each element to ensure
continuity at each node.
 Imposing boundary conditions
 We impose the necessary boundary
conditions at the boundary nodes
 Solving global system of equations Coarse-grained within a Finite
Element mesh.
J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

12
 13/10/2015

1.4 Hybrid QM/MM methods

 Which is the main limitation to
the classical force fields ? Inner Region

 Breaking/Forming bonds
 Complex interactions beyond the
classical model
 Proposed solution (QM/MM):
 Split the system in two parts
 Inner Region  High level description
 Quantum Mechanics (QM)
 Outer Region  Low level description
 Molecular Mechanics (MM)
Metal-Induced Chiral Poly-thiophene Aggregate

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

1.4 Hybrid QM/MM methods

 Energy partition
E(S )  EQM (I )  EMM (O)  EQM MM (I , O)

EQM  MM ( I , O )  EQM
vdw
 MM ( I , O )  EQM  MM ( I , O )  EQM  MM ( I , O )
el bonded

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

13
 13/10/2015

1.4 Hybrid QM/MM methods

 Environment effect (Electrostatics)
 Mechanical Embedding
 QM no influenced by environment
 QM...MM electrostatic interactions
 handled using point charges on QM
atoms
 Electrostatic Embedding
 QM calculation is polarized by point
charges on MM atoms
 MM atoms have a force arising from
their interaction with QM
Warshel, Nobel Lecture (2013)
 Some Implementation Programs
 CPMD (Car Parrinello based) http://www.cpmd.org/
 Amber ( QM/MM-MD) http://ambermd.org/
 PUPIL ( QM/MM-MD ) multiple MD and QM packages http://pupil.sourceforge.net/
 ChemShell (QM/MM) multiple MM engines and QM packages
http://www.cse.scitech.ac.uk/ccg/software/chemshell/manual/index.html

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

1.4 Hybrid QM/MM methods

 QM/MM – Molecular dynamics
 Atomic motions handled by molecular dynamics (MD Engine)
 Similarly to the QM/MM approach:
 Energies and forces are calculated by dividing the system into two
different parts.
 Region of interest treated at the quantum mechanical level (QM Engine)
 Molecular mechanics (MM) are applied to the rest of the system by using a
classical potential energy function.
 At each MD step MM and QM energies and forces are merged to
get next nuclear coordinates of the system.

MD Engine QM/MM QM Engine

coupling

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

14
 13/10/2015

1.4 Hybrid QM/MM methods

 QM/MM – Molecular
T T T T T T T
dynamics
Gaussian Siesta deMon2K NWChem MNDO Amber DL_POLY

 PUPIL: A modular
IGaussian ISiesta IdeMon2K INWChem IMndo IAmber IDlPoly
implementation
GENERAL PUPIL INTERFACE OOP  Independent packages
T
running together
OOP  Manager conduct multi-
PUPIL
S
step simulation
MANAGER
 GUI helps to prepare
the simulation
S  Server
C  Client DB  Manager and GUI are
T  Thread asynchronous
C OOP  I/O data exchanged
PUPIL
GUI
through database

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

1.4 Hybrid QM/MM methods

 QM/MM – Molecular dynamics
 Reactivity: Diels-Alder reaction (water solvent)

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

15
 13/10/2015

1.5 The Simulation Concept. Simulation

Types
 Simulation types
 1.5.1 Stochastic: Monte Carlo Method
 Systems with a large number of coupled degrees of freedom
 A random exploration of conformational space.

 1.5.2 Mechanical: molecular dynamics, Brownian dynamics

 Temperature (particle velocities)
 Conformational entropy

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

1.5.1 Stochastic: Monte Carlo Method

 Is a statistical evaluation of mathematical functions using
random samples.
 Systems with a large number of coupled degrees of freedom
 A random exploration of conformational space.
 Requires a good source of random numbers
 Generation of random conformations
 Metropolis algorithm allows a clever selections of states.
 Initially formulated for the equation of state of hard spheres

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

16
 13/10/2015

1.5.1 Stochastic: Monte Carlo Method

 The error involved in this scheme is mitigated as much number of random
samples are taken
 The term “Monte Carlo” was apparently first used by Ulam and von
Neumann as a Los Alamos code word for the stochastic simulations
 Monte Carlo Algorithm
Generate initial structure R. Calculate V(R). n  0
do
Modify structure: R’. Calculate V’(R’)
If V’<V then
R  R’ ; V  V’
else
Generate random number RND
If exp(-(V’-V)/kT) > RND
R  R’ ; V  V’
endIf
endIf
n n+1
while ( n < N)

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

1.5.2 Mechanical: molecular dynamics,

Brownian dynamics
 Molecular dynamics
 The particle positions under an specific interatomic potential (force
field) are updated every short increment of time.
 Atomic trajectories of a system of particles are obtained by
numerical integration of Newton’s equation of motion,
 Initial conditions and boundary condition have to be specified

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

17
 13/10/2015

1.5.2 Mechanical: molecular dynamics,

Brownian dynamics
 Molecular dynamics algorithm

Give atoms initial positions R(t=0). Choose short ∆t. n  0

do
Get forces:
Apply thermostat and volume changes
Move atoms: 
∆
1
while ( n < N)
Analyze the data

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

1.5.2 Mechanical: molecular dynamics,

Brownian dynamics
 Dynamic of proteins
 Structural study of proteins
 Protein-ligand study, etc.
 Dynamics of advanced materials
 Structure and thermodynamics
 Extreme conditions and properties
 Depositions, etc.

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

18
 13/10/2015

1.5.2 Mechanical: molecular dynamics,

Brownian dynamics
 Brownian dynamics
 Diffusional processes are important in many
systems
 Brownian dynamics provides an interesting
description of diffusional processes
 Brownian dynamics can access longer
timescales and spatial scales than molecular
dynamics
 Applications of BD simulations Brownian motion simulation of a big particle
(dust particle) that collides with a large set of
 Protein Folding smaller particles (molecules of a gas)
https://en.wikipedia.org/wiki/Brownian_motion
 DNA Bending and Supercoiling
 Enzyme-Substrate Encounter
 Protein-Protein Encounter
 Diffusion in the Cell

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

1.5.2 Mechanical: molecular dynamics,

Brownian dynamics
 Brownian dynamics permits simulation of macromolecular motions on the
millisecond time scale
 keeping atomic level accuracy in the representation of the molecules
 usually neglecting the internal dynamics of macromolecules.
 To simulate large systems over longer times, it is necessary to develop reduced
models.
 e.g., one protein domain is represented by one sphere
 Interaction properties are derived from the (non-spherical) atomic-detail structure of the
domain.
 MD and BD simulations are comparable in the time scale of 1-100 ns
 BD and RM simulations are comparable in 1-10000 ns

J.Torras - UPC (2015) Molecular Modelling - Computational Modelling in Physics, Chemistry and Biochemistry

19
 13/10/2015

Computational Modelling in Physics,

Chemistry and Biochemistry

Molecular Modelling

Unit 1. Multi-scale Description of Atomic and Molecular

Systems

J. Torras , Univ. Politècnica Catalunya (2015)

20