Molecular modelling in drug design

Laboratorio di Progettazione dei Farmaci

Dipartimento di Farmacia
Università degli Studi di Parma

Suggested readings

!  An introduction to medicinal chemistry

!  G.L. Patrick
! Chs. 13 and 15 (QSAR and modelling, introductory)
! Oxford Univ. Press, 2005 (3rd edition)

!  Chemoinformatica
!  M. Mabilia
! Springer-Verlag Italia, 2012

!  Essentials of Computational Chemistry

!  C.J. Cramer
! Wiley, 2004 (2nd edition)

Laboratorio di Progettazione dei Farmaci 1

Molecular models in drug design: how they work
!  Energy
!  E = f (geometry)

!  Model approximations
!  Approximated energy functions
!  Neglect of boundary conditions
!  Neglect of entropic terms

!  The level of approximation must be tuned to provide useful information:

!  "Stable" geometries
! Drug structure "cleaning"
! Protein structure refinement
! Drug allocation within a binding site
!  Probability ranking
! Most populated conformation of a drug
! Prediction of drug-target affinity

Laboratorio di Progettazione dei Farmaci 3

Static or dynamic representation of molecular properties

!  Static representations are allowed for:
!  One or more representative minimum-
energy conformations
! Accessible conformations in the
Active Analog Approach
!  Average properties based on
experimental data (empirical)
! Surface lipophilicity from LogP data

!  Dynamic representations are made

possible by:
!  Conformational space exploration
! <A> = Σi Ai·Pi
!  Trajectory representation in MD
simulation

Laboratorio di Progettazione dei Farmaci 4

Energy: the pivotal concept in molecular modelling
!  Probability of states: the Boltzmann distribution
!  NA/NB = gA/gB · e -ΔE/RT
! When T = 25°C:
! ΔE = 1 Kcal/mol NA/NB = 1 / 5.4
! ΔE = 5 Kcal/mol NA/NB = 1 / 4650
! ΔE = 10 Kcal/mol NA/NB = 1 / 21,600,000

!  Force = energy derivative

!  Static equilibrium in minimum-energy states

!  Chemical equilibrium and free energy

!  A ! B Keq = e -ΔG°/RT
!  ΔG° = ΔH° - T ΔS° H = E + pV

!  Chemical transformations and activation energy

!  A → B k = A e -Ea/RT

Laboratorio di Progettazione dei Farmaci 5

Energy functions as models

!  E = f (x1, x2 … xK; β1, β2, … βM) + ε
!  Within the application domain, the function f(x) is the "best"
approximation of the "true" energy, for the geometry described by the
variables x = {x1, x2 … xK}

!  Two kinds of models:

!  Mechanistic models
! Known relationship between E and x
! Known f and β
! Model bias from inappropriate (approximated) f and β
! Estimates of βm : semi-empirical models
!  Empirical models
! Unknown f
! Select relevant xj
! Choose appropriate model function and complexity
! Estimate βm
! Check model validity
!  For empirical and semi-empirical models: calibration
! Fitting of calculated properties to known ones.
! Model bias and uncertainty due to inappropriate model calibration
Laboratorio di Progettazione dei Farmaci 6
Energy functions: quantum mechanics
!  Solving electronic Schrödinger equation: the wavefunction
HΨe = EΨe

!  The Hamiltonian operator

!  Accounts for kinetic and potential energy of each electron
!  Approximations:
! Born-Oppenheimer approximation
!  Electronic and nuclear motion separated:
!  electrons move in the electric field created by fixed nuclei
!  Ψ and E are calculated (iteratively) for a fixed molecular geometry
! Hartree-Fock approximation
!  Electron's Ψ is calculated in the average static field of other electrons
!  Electron correlation cannot be neglected in many cases
! Molecular orbital
!  Approximated as combination of atomic orbitals (AOs)
!  AOs approximated by combination of primitive functions (basis sets)

!  Semi-empirical methods (MNDO, AM1, PM3….)

!  Neglect of overlap integrals
!  Introduction of empirical parameters

Laboratorio di Progettazione dei Farmaci 7

Energy in molecular models

!  For static models of isolated systems:
Energy landscape
!  E: potential energy
Landscape
! By quantum mechanics, molecular mechanics, for φ/ψ plane
semiempirical of dialanine
or hybrid Comp
methods
!  For different geometries (states)
! E = f (geometry) " P.E.S. E

!  For “real” systems

ψ
φ
!  Molecular partition function
! Q = Σ exp(-E/kT)
! Sum extended to all states (discrete)
!  Free energy (Helmholtz):
! A1-A0 = -kT ln(Q1/Q0)

!  Dynamic and thermodinamic properties of a real system depend on the

P.E.S., describing ALL accessible states
Molecular
Laboratorio di Progettazione dei Farmaci Dynamics Simulation - Michel Cuendet - EMBL82008
Molecular mechanics
!  Molecules: interacting objects described by
simplified energy functions
!  Direct (not iterative) calculation of E
!  Quick calculation of E for different geometries

!  Atoms (nuclei)
! Different atom types (hybridization)
! Radius (VdW) to decribe attractive
(dispersion) and repulsive forces
! Partial charges to describe electrostatic
interactions
! Pseudo-atoms (lone pairs) possible

!  Bonds (molecular topology)

! Different bond types
! consistent with the couple of linked atoms

!  Atom types, partial charges and bonds generally

remain unchanged

Laboratorio di Progettazione dei Farmaci 9

The force field

!  The energy of each atom depends on the force field created by other
atoms, through:
!  bonded interactions: from atoms linked through:
! 1 bond: stretching
! 2 bonds: bending
! 3 bonds : torsion (sometimes considered as a non-bonded interaction)
!  non-bonded inteactions: from atoms not directly linked
! Van der Waals
! Electrostatic
!  correction terms: to improve the accuracy of energy estimates

!  The internal energy (steric energy, conformational energy) of the

system is the sum of all the contributions (terms)
!  All atoms
!  All interactions

!  Electron state is not defined, but it is implicitly considered in the force

field terms

Laboratorio di Progettazione dei Farmaci 10

Energy functions in molecular mechanics
!  Additive empirical function (with correction factors):
20
Estr quadratic
E str anharmonic
E = ΣEstr,ij + ΣEbend,ijk + ΣEtor,ijkl + ΣEVdW,ji + ΣEVdW,ji + ...

E str (kcal/mol)
10

!  Bonded interactions
!  Stretching
! Estr,ij = 1/2 k_sij ( rij - ro )2 +1/2 k_anij ( rij - ro )3 0

!  Bending 0.0 1.0 r

o 2.0
r (Å)
3.0

! Ebend,ijk = 1/2 k_bijk ( θijk - θο )2

!  Torsion
! Etor,ijkl = 1/2 k_tor1 (1-cos φ) +1/2 k_tor2 (1-cos 2φ) + 1/2 k_tor3 (1-cos 3φ)
!  Out-of-plane -φ +φ
ri j
j A
C B
!  Non bonded interaction i
θijk
k

!  Van der Waals

! EVdW,ji = - A/rij6 +B/rij12
!  Electrostatic
! Eelec,ij = qiqj/4πεrij

Laboratorio di Progettazione dei Farmaci 12

Force field parameterization and use

!  Parameters are obtained from experimental data, derived from different sources,
such as:
!  X-ray diffraction, electronic diffraction
! Bond lenght and angle wideness
!  Spectroscopy
! Bond stretching constant
!  "Ab initio" calculations

!  Force fields are parameterized to give:

!  Good geometries
!  Relative conformational energies

!  Force field selection

!  FF for specific (macro)molecules
! AMBER, GROMOS, CHARMM, GLYCAM
!  FF for small molecule modelling
! MM2, MM3
!  Universal FF
! Tripos, MMFF94, OPLS-X

Laboratorio di Progettazione dei Farmaci 13

Energy and geometry: the informatics
!  Ecalc = f (x1, x2 … xK; β1, β2, … βM)
Z - Matrix:
1 C
!  Degrees of freedom for internal 2
3
O
C
1
1
1.42100
1.51400 2 107.4000
energy 4
5
H
H
1
1
1.11098
1.11098
2
2
106.6996
106.6996
3
3
109.4100
109.4100
!  3N-6 6
7
H
H
2
3
0.94193
1.11298
1
1
106.8988
109.9993
3
2
179.9802
180.0000
8 H 3 1.11298 1 109.9989 7 109.0010
!  3N-5 for linear molecules 9 H 3 1.11298 1 109.9989 7 109.0010

PDB File:
!  Cartesian coordinates HETATM
HETATM
16
17
N
CA
DAL
DAL
A
A
12
12
-2.846
-2.252
4.674
4.096
22.172
23.404
!  Internal coordinates HETATM
HETATM
18
19
CB
C
DAL
DAL
A
A
12
12
-3.211
-2.065
2.999
5.256
23.902
24.408
HETATM 20 O DAL A 12 -2.639 6.325 24.244
HETATM 21 N DAL A 13 -1.261 5.054 25.442
HETATM 22 CA DAL A 13 -1.000 6.145 26.394
HETATM 23 CB DAL A 13 0.094 7.130 25.925
HETATM 24 C DAL A 13 -0.635 5.531 27.776
!  System configuration (input): HETATM
...
25 O DAL A 13 -0.690 6.281 28.778

CONECT 16 17
!  Quantum mechanics (QM) CONECT 17 16 18 19
CONECT 18 17
! Atoms, charge, x,y,z. CONECT 19 17 20
CONECT 20 19
!  Molecular mechanics (MM) CONECT 21 22
CONECT 22 21 23 24
! Atom types, connectivity, x,y,z. CONECT
CONECT
23
24
22
22 25 26
CONECT 25 24

Laboratorio di Progettazione dei Farmaci 14

System configuration (input) for molecular mechanics

H
H
H
!  Atom types C.3
H
am
!  Bond types N.am
C.am

!  Protonation state H
1 O.2
H C.3
H
!  Assignment of partial charges H
H C.ar H

ar 2 C.2
!  After a structure has been drawn through
graphical interface, you must check: H
N.pl3

!  Atom types H
H
!  Bond types
!  Protonation
!  Tautomerism
!  Partial charges
!  Starting geometries
!  Parameter availability and quality
! C.2-C.3-C.3-N.am: k_tor1=0.0
k_tor2=0.0 k_tor3=0.7 quality=low
Laboratorio di Progettazione dei Farmaci 15
Charge distribution
Charge distribution is important for intra- and intermolecular propeties
nuclei electrons
!  Dipole moment µx = ∑Z k ⋅ xk − ∑ ∫ψ (r' ) ⋅ x ψ (r' )dr'
k
k j

!  Polarizability α = ∂µ x ∂Ex (electric field strength )

!  Molecular electrostatic potential (MEP)

!  Potential energy at position r:
nuclei
Zk 1
VMEP (r ) = ∑ r − r − ∫ Ψ (r' ) r − r' Ψ (r' )dr'
k k

! Ψ depends on
! nuclei position (Rk) -> geometry
! External fields (hamiltonian)

!  Point charge (mulipole) approximation: Electronic charge distribution for formaldehyde.

nuclei Lines represent isopotential surfaces.
qk Dots: VdW surface
VMEP (r ) ≈ ∑ k r − rk
Laboratorio di Progettazione dei Farmaci 16

Energy minimization - Geometry optimization

!  Energy minimization is an iterative process.
!  Coordinates are gradually changed to produce conformations with lower
energy, until the minimum is reached.
!  Convergence if grad E = 0; ∇2 E > 0

!  Local and global minima

!  The starting point influences the solution found.

!  Geometry optimization:
!  By multivariate optimization of E = f (X)
! X: cartesian or internal energy
! f(X) has 3N-6 df, but more variables are often employed

Laboratorio di Progettazione dei Farmaci 21

Energy minimization methods
!  Iterative methods I: Function
II: First derivative
!  Starting geometry III: Second derivative
!  Geometry changes based on
! Energy, force, PES curvature
!  Convergence
! Stop when two steps have very close:
!  Energy difference
!  Change in coordinates
!  Root-mean square gradient

!  Non-derivative methods
!  Simplex

!  First order minimization methods:

!  gradient vector g
! Steepest descent
! Conjugate gradients

!  Second derivative methods:

!  Hessian matrix H
! Newton-Raphson
Laboratorio di Progettazione dei Farmaci 22

SBDD: interaction with ligands

!  Interaction free energy
!  Enthalpy
! from representative structures
!  Entropy
! Complete exploration of conformational space
! Empirical estimations

!  Approximated models
!  Approximated energy functions
!  Conformational space sampling
!  Neglect of environmenal effects

!  Docking
!  Complementarity
!  Different poses

!  Scoring functions
!  Energy based
!  Knowledge based
!  Empirical

Laboratorio di Progettazione dei Farmaci 36

Docking
!  Ligand/protein conformations

!  Based on stereoelectronic complementarity

!  Alternative poses O

N O CONH 2
H

Gln273

Ser241 Ser241
Leu192
Leu380

Thr488

A B

Laboratorio di Progettazione dei Farmaci 37

Scoring functions
!  Energy based
!  Knowledge based
!  Empirical

!  Virtual Screening
!  Compound ranking

Laboratorio di Progettazione dei Farmaci 38

LBDD: Pharmacophore analysis

!  On a class of ligands sharing the same binding site, with the same
mechanism

!  Putative pharmacophore elements

!  Charged atoms (+ or -)
!  HB acceptors or donors
!  Lipophylic centers

!  Accessible conformations

!  Pharmacophore model building

!  Elements
!  Distances (tolerance)

Laboratorio di Progettazione dei Farmaci 39

Active analog approach – Pharmacophore analysis

!  1. select common pharmacophore elements
!  2. build all accessible conformations
!  Calculate distances among pharmacophore elements
!  3. select only elements and distances common to at least 1 accessible
conformation of all compounds
!  Tolerance
!  Superposition of “active” conformations: accessible space

H CH3
dO-H N
H3C dO-H
O O
dO-O H CH3
O
H3C N
O
N dO-O
H
modelli
composto A composto B farmacoforici

1
dO-H dO-H dO-H
∩ = 2

dO-O dO-O dO-O

Laboratorio di Progettazione dei Farmaci 40
Pharmacophore model for MT2 selective antagonists
NHCOCH 2CH3 NHCOCH 3 O
NHCOCH 3 NHCOCH 3

N
N O
H
Cl

4P-PDOT luzindole

!  Pharmacophore elements Superposition

Laboratorio di Progettazione dei Farmaci 41

LBDD: 3D-QSAR

For each molecule, Interaction fields Y X

aligned following
a pharmacophore

PLS

pseudo-coefficients

color

Laboratorio di Progettazione dei Farmaci 44

3D-Q Structure-Affinity Relationships

MT1 MT2
SDy=0.51 n=34, 3 L.V. SDy=0.85 n=34, 5 L.V.
q2=0.36, SDEP=0.43 q2=0.67, SDEP=0.53
R2=0.69, s=0.30 positive R2=0.90, s=0.29
negative

Laboratorio di Progettazione dei Farmaci 45

3D-QS-X Relationships

Selectivity Intrinsic activity

positive
negative
MT1
MT2

[Rivara S. et al., J. Med. Chem., 2003, 46, 1429 ]

Laboratorio di Progettazione dei Farmaci 46