Conformational Analysis

3D Structures, Conformations and Molecular Surfaces

Christof H. Schwab

Molecular Networks GmbH Altamira LLC

Henkestr. 91 1455 Candlewood Drive
91052 Erlangen, Germany mn-am.com Columbus, Ohio 43235, USA
Molecular Networks and Altamira
MN-AM

Erlangen, Germany Columbus, Ohio, USA

Friedrich-Alexander-Universität The Ohio State University
1997 2008

 Chemoinformatics
 3D structure generation
 Physicochemical and reaction properties
 Metabolic reaction knowledge
 Computational toxicology and risk assessment
 Database and knowledgebase
 Predictive models
 Consulting services

2
Product Lines – Chemoinformatics

 CORINA Classic
CORINA Classic
 Industry-standard 3D structure generation
 CORINA Symphony
 Profiling and managing of chemical datasets
 Workflows
 Structure cleaning/processing
 Descriptor generation (properties and fragments)
 SYLVIA
 Estimation of synthetic accessibility of compounds
 Public tools
 ChemoTyper (with ToxPrint Chemotypes)
 https://chemotyper.org, https://toxprint.org
3
Product Lines – Computational
Toxicology and Risk Assessment
 ChemTunes
 Platform to support decision making in human health
and regulatory critical endpoints
 Toxicity database (all endpoints)
 "Inventory" concept for compound location
 ChemTunes ToxGPS
 Prediction models
 All human health related endpoints
 Workflows
 TTC (thresholds of toxicological concern)
 Read-Across (in development)
Genotoxic
Impurities
Workflow

 ICH M7 GTI (in development)

4
3D Structures, Conformations and
Molecular Surfaces – Overview
 3D structures – why, where needed and how many
 3D structure generation
 Single 3D structures and conformational ensembles
 Methods, approaches, applications
 Storage of 3D structures
 Examples of file formats
 Molecular surfaces
 Approaches, application

 References

5
3D Structures – For What?

 Chemists' language
 2D structure diagrams
 Stereochemistry
 Molecules are 3-dimensional objects
 Spatial interactions of molecules cause effects

6
3D Structures – Where needed?

 3D database generation
 Lead discovery and lead optimization
 Ligand- and structure-based virtual screening
 Docking studies, pharmacophore and similarity searching
 Prediction of chemical, physicochemical and biological properties
 ADMET properties
 QSAR and QSPR studies
 Structure elucidation
 Prediction of chemical reactivity
 Input to quantum-mechanical and force field calculations

7
Chemical Structures – How many?

 The Cambridge Crystallographic Database (CSD)

 Over 850,000 experimentally-determined crystal structures
 PubChem
 82 Million characterized chemical compounds
 Chemical Abstract Service, CAS Registry
 121 Million unique organic and inorganic chemical substances
 GDB, University of Berne
 GDB-13, 1 Billion compounds (13 atoms, C, O, N, S, and Cl)
 GDB-17, 164 Billion compounds (17 atoms, C, O, N, S, and halogens)
 RCSB Protein Data Bank, PDB
 123,000 biological macromolecular structures
 21,000 ligand structures
8
Automatic 3D Structure Generators

 Categories of 3D molecular model building approaches

Fragment-based Rule- and data-based Automatic

methods methods 3D model
builders

Numerical methods

 No sharp borders between approaches

 Fragment-based methods require rules
 Rule- and data-based methods use 3D templates
 Numerical methods require starting geometries

9
3D Structure Generators

 Fragment-based methods
 Fragment, select appropriate 3D template, link
 Minimum set of rules for fragmentation, analog searching and linking
 Rule- and data-based methods
 Knowledge base from theoretical investigations and experimental structures
 Explicit rules and implicit data
 Numerical methods
 Molecular force field calculations
 Quantum-mechanical methods
 Semi-empirical, DFT and ab initio
 Distance geometry

10
3D Structure Generators – General
Issues (1/3)
 Different conformational behavior of cyclic and acyclic portions

 Ring systems with limited degrees of 56 kJ/mol 25 kJ/mol

freedom 50 kJ/mol 25 kJ/mol

 Use of "allowed" low-energy ring templates
27 kJ/mol 18 kJ/mol

 Open-chain structures with increasing

number of degrees of freedom with
increasing number of rotatable bonds
 Selection of low-energy conformation 𝑛
360
 Principle of longest pathways 𝑁=
𝑘

11
3D Structure Generators – General
Issues (2/3)
 Stereochemistry awareness
 Tetrahedral chiral centers and (R) (S)
cis/trans double bonds
 Coded in input structure

 Close contacts and steric crowding

 Mechanisms to eliminate such situations

 Conformational analysis
 Several conformations to identify a low-energy one

12
3D Structure Generators – General
Issues (3/3)
 Processing of large amounts of data
 Millions of structures in (company) databases
 Robustness
 CPU times
 High conversion rates
 Handling of broad range of chemical spaces

 Quality of 3D models
 Close contacts and steric crowding
 Reproduction of experimentally determined geometries

13
Fragment-based Methods

 Fragment the structure

N,N-dimethyl-benzamide
 Rules for fragmentation
 Select appropriate 3D template
 Library of 3D templates
 Link 3D templates
 Rules for linking

 Rules and data required

Conjugated
Phenyl Amide N-methyl
ketone

14
Rule and Data-Based Methods

 Knowledge base from theoretical investigations and experimental

structures
 Explicit and implicit rules and data
 Standard bond lengths and angles
 Preferred torsion angles
 Ring templates
 Empirical (fast) energy estimation
 Increments for ring templates

CORINA Classic

15
CORINA Classic – General Principles

Internal coordinates
stereo information Tables

Connection table Stereo information

Stereo descriptors

Fragmentation
Small to medium-
sized rings
Acyclic Ring
portions systems
Large/macrocyclic
3D model rings

Re-assembling
Geometry
conformational analysis
optimization
(close contacts)
16
CORINA Classic

 Atom types and bond angles

tetrahedral trigonal bipyramidal octahedral

linear: 180o planar: 120o
109.47o 90o, 120o, 180o 90o, 180o

 Ring templates
 Smallest set of smallest rings

+


+
+
17
CORINA Classic

 Macrocyclic structures
 Principle of superstructure

N N
O H H

O O

S S

2D structure Superstructure 3D model

18
CORINA Classic – Application

 Conversion of National Cancer Institute (NCI) database

 Number of structures 265,242
 Structures converted 263,184
 Structure coding errors in DB 543
 Conversion rate 99.43%
 CPU time 0.5 h
(x86 Linux, 1.5 GHz) (1,843 s, 0.007 s/molecule)

19
CORINA Classic – Key Features

 Support of a variety of chemical file formats

 SDF, RDF, SMILES, PDB, SYBYL MOL and MOL2, MacroModel, Maestro, CIF,...
 Generation of multiple ring conformations
 Data-based for rings consisting of up to nine atoms
 Interface to docking program FlexX
 Generation of stereo isomers
 Automatic detection of stereo centers (tetrahedral and cis/trans)
 Full/partial enumeration, preserve defined centers, duplicate detection
 Structure "clean-up" features
 Adding H atoms, neutralizing formal charges, atom and bond type assignment,
stereo information, removal of counter ions in salts, ...

20
CORINA Classic Interfaces

 Command line interface and

library version
 Linux shared object and Windows DLL

 Pipeline Pilot component

 KNIME node

21
Numerical Methods

 Molecular force field calculations

 Quantum-mechanical methods
 Semi-empirical
 DFT
 ab initio
 (Distance geometry)

22
Molecular Force Field Calculations

 Classical mechanical treatment of molecules k

 Composed of masses: atoms
m1 m2
 Connected by springs: bonds
 Potential energy by Hooke's law
r
𝑘
 Born-Oppenheimer approximation 𝑉(𝑟) = (𝑟 − 𝑟0)
2
 Separation of movement of electrons from much slower
movement of nuclei
 Potential energy of molecule as function of atomic coordinates

23
Molecular Force Field Calculations

 Two major parts

 Functional form (mathematics)
 Parametrization and atom typing

 Not a single, "true" mathematical expression

 Different functional forms can be applied
 Different sets of parameters and atom types
 Derived from experimental results
 Represent thermodynamic average (rather than a particular geometry)

 Empirical method
24
Molecular Force Field Calculations

 Functional form
 Calculation of potential energy
 Optimization of energy depending on atom coordinates
 Bonded and non-bonded contributions

𝑉= 𝑉𝑏𝑜𝑛𝑑𝑠 + 𝑉𝑎𝑛𝑔𝑙𝑒𝑠 + 𝑉𝑡𝑜𝑟𝑠𝑖𝑜𝑛𝑠 + 𝑉𝑒𝑙𝑒𝑐𝑡𝑟𝑜𝑠𝑡𝑎𝑡𝑖𝑐 + 𝑉𝑣𝑎𝑛𝑑𝑒𝑟𝑊𝑎𝑎𝑙𝑠

d+ d-

Bond Angle Electrostatic van der Waals

Torsion
stretch bend (non-bonded) (non-bonded)

25
Molecular Force Field Calculations

 Atom typing
 Differentiation between hybridization state, local environment, special
conditions (e.g., strained ring)
 Depending on application

 Parametrization
 The more atom types, the more parameters for contributions/terms in energy
function required

𝑉= 𝑉𝑏𝑜𝑛𝑑𝑠 + 𝑉𝑎𝑛𝑔𝑙𝑒𝑠 + 𝑉𝑡𝑜𝑟𝑠𝑖𝑜𝑛𝑠 + 𝑉𝑒𝑙𝑒𝑐𝑡𝑟𝑜𝑠𝑡𝑎𝑡𝑖𝑐 + 𝑉𝑣𝑎𝑛𝑑𝑒𝑟𝑊𝑎𝑎𝑙𝑠

26
Force Fields – Applications

 Conformational analysis
 Studying different conformations of a molecule
 Structure refinement using experimental data
 Molecular properties
 Heats of formation

 Used in molecular dynamics simulations

27
Some Force Fields and Programs
(1/2)
 MM2/MM3/MM4 (Allinger, University of Georgia)
 Small molecules, peptides, proteins
 UFF (Rappé, Goddard III, Colorado State University)
 Mainly small molecules, generic force field for "whole periodic table"
 AMBER (Kollman, University of California San Francisco)
 Biomolecules including solvation effects (water models TIP3/4P, SPC, POL3,…)
 CHARMM Karplus, Harvard University)
 Macromolecular simulations, molecular dynamics
 GROMOS (van Gunsteren, Berendsen, University of Groningen)
 Molecular dynamics, aqueous or apolar solutions, lipid systems

28
Some Force Fields and Programs
(2/2)
 MMFF (Halgren, Merck and Co, Inc)
 Optimized to handle functional groups relevant in pharma

 Vendors
 BIOVIA (Accelrys)
 Chemical Computing Group
 Schrödinger

 Free/open source
 Open Babel
 RDKit

29
Quantum Mechanical Approaches

 Solving the molecular Schrödinger equation

 ĤY = EY
 Energy as a function of the electronic wave function

 Three main categories according to level of theory

 Semi-empirical molecular orbital (MO) theory
 Density functional theory
 Ab initio MO theory

 Computationally expensive
 Semi-empirical < DFT < ab initio

30
Quantum Mechanical Approaches

 General principles
 Electronic structure of molecule (built by atoms) as basis
 Linear combination of atomic orbitals (LCAO) approximation to derive molecular
orbital
 Hartree-Fock (HF) and self-consistent field (SCF) approximation
 Each electron interacts with mean field of all other electrons

 Semi-empirical methods
 Parametrization of (some) terms using data from experimental data
 Neglecting/approximation of (some) interaction terms

31
Output of Quantum Mechanical
Calculations
 Net atomic charges
 Dipole and higher multipole moments
 Polarizabilities
 Orbital energies Valuable for modelling
 HOMO/LUMO of chemical reactivity
and toxicity endpoints
 Heat of formation Molecule orbitals of phenylalanine
 Surface descriptors (HOMO/LUMO)

 Local ionization potential

 Atomic coordinates

32
Quantum Mechanical Program
Packages
 GAUSSIAN
 Molecular mechanics, semi-empirical methods (AM1, PM3, …), ab initio
 MOPAC
 Semi-empirical methods, AM1, PM3, …
 Spartan
 Molecular mechanics, semi-empirical methods, ab initio models, DFT, …
 Hyperchem
 Molecular mechanics, molecular dynamics, semi-empirical and ab initio methods
 EMPIRE
 Semi-empirical methods

33
Conformational Sampling

 Generate set of diverse conformations

 Biologically relevant conformations in drug discovery
 Bioactive conformation
 Conformation in receptor-bound state

 General case of 3D structure

generation

 Applications, e.g.,
 Docking
 Pharmacophore searching

34
Conformational Sampling – General
Workflow
 Identification of rotatable bonds
 Flexible ring systems
 Open-chain portions

 Generation of conformations
 Application of implemented algorithm
 Avoidance of "un-favorable" conformations
 Combinatorial explosion

𝑛
360
𝑁=
𝑘

35
Conformational Sampling – General
Workflow
 Checking for duplicates and very similar conformations
 Inter-conformational distances in Cartesian space 𝑁 2
𝑖 ∆𝑑
 Root mean square deviation (RMSDXYZ) 𝑅𝑀𝑆𝐷 =
𝑁

 Selection of a set of representative conformations

 Subsampling
 E.g., clustering in Cartesian space
 Ideally done in one of previous steps to bias search towards area of application

36
Conformational Sampling – Methods

 Rule- and data-based, fragment-based methods

 Empirical and fast

 Systematic searches
 Exhaustive, e.g., grid searches
 Combinatorial explosion

 Random methods
 Random change of Cartesian or
internal coordinates
 Subsequent optimization

37
Conformational Sampling – Methods

 Genetic algorithms
 Robust optimizers
 Torsion angles as genes
 Mutation/crossover operators

 Simulation methods
 Molecular dynamics
 Simulated annealing

38
Conformer Generator ROTATE Classic

 Hybrid approach
 Systematic search
 Rule- and data-based system

Torsion angle distributions in Empirical energy function

small molecule crystal structures 𝐸 𝜏 = −𝐴 ln 𝑓(𝜏)
39
Conformer Generator ROTATE Classic

 Superimpositions of experimentally determined receptor-bound

geometries and ROTATE-generated models with smallest RMSXYZ

2L-benzylsuccinate (3cbx)
RMSXYZ = 0.27 Å

DMP 450 (1dmp) Citric acid (3cts)

VX-478 (1hpv) RMSXYZ = 0.34 Å
RMSXYZ = 1.59 Å
RMSXYZ = 0.94 Å

40
Available Methods

 CatConf/ConFirm  MOE
 Accelrys/Biovia  Chemical Computing Group
 CEASAR  MacroModel/ConfGen
 Accelrys/Biovia  Schrödinger
 CORINA Classic and
ROTATE Classic  Free/open source
 Molecular Networks  Open Babel
 OMEGA  RDKit
 OpenEye
 CONCORD/CONFORT
 Tripos/Certara

41
Storage of 3D Structures

 Various standard chemical file formats

 Cartesian (x,y,z) coordinates

 Molfile, SD file
 SYBYL MOL/MOL2
 PDB

 Internal coordinates
 Crystallographic file formats, e.g., CIF
 (Z matrix)

42
Storage of 3D Structures

 Cartesian coordinates  Internal coordinates

y
y 6 6
z
y 7 5 r1 5 7
H2
H1 1 2 12
C1 a 4 8 x 8 4 z
r2
H3 H4 t4-2-1-3
z 3 3
C1
x C2 1.5 1
x y z Cl3 1.7 1 109 2
C1 -0.0127 1.0858 0.0080 H4 1.1 2 109 1 -60 3
H1 0.0021 -0.0041 0.0020 H5 1.1 1 109 2 180 4
H2 1.0099 1.4631 0.0003 Cl6 1.7 2 109 1 60 5
H3 -0.5399 1.4469 -0.8751 H7 1.1 1 109 2 -60 6
H4 -0.5229 1.4373 0.9048 H8 1.1 2 109 1 180 7

43
Molfile, Structure Data (SD) File

 SD V2000
L-alanine
10191614583D 1 1.00000 0.00000

7 6 0 0 0 0 0 0 0 0999 V2000
-0.0184 1.5028 0.0103 C 0 0 0
0.0021 -0.0041 0.0020 C 0 0 0
-0.7002 -0.5305 1.2552 C 0 0 0
1.0197 2.1211 0.0037 O 0 0 0
-1.1898 2.1580 0.0194 O 0 0 0
1.3935 -0.4748 -0.0138 N 0 0 0
-0.5153 -0.3696 -0.8850 H 0 0 0
1 2 1 0 0 0 0
2 3 1 0 0 0 0
1 4 2 0 0 0 0
1 5 1 0 0 0 0
2 6 1 0 0 0 0
2 7 1 1 0 0 0
M END

44
SYBYL MOL/MOL2 File

 SYBYL MOL2 file with atom typing

@<TRIPOS>MOLECULE
L-alanine
7 6 0 0 0
SMALL
NO_CHARGES
@<TRIPOS>ATOM
1 C1 -0.0184 1.5028 0.0103 C.2
2 C2 0.0021 -0.0041 0.0020 C.3
3 C3 -0.7002 -0.5305 1.2552 C.3
4 O4 1.0197 2.1211 0.0037 O.2
5 O5 -1.1898 2.1580 0.0194 O.3
6 N6 1.3935 -0.4748 -0.0138 N.3
7 H7 -0.5153 -0.3696 -0.8850 H
@<TRIPOS>BOND
1 1 2 1
2 1 4 2
3 1 5 1
4 2 3 1
5 2 6 1
6 2 7 1
# End of record

45
(Brookhaven) Protein Data Bank File

 Example
HEADER UNK 16-10-19 1UNK
COMPND L-alanine
REMARK
HETATM 1 C1 UNK 1 -0.018 1.503 0.010 1.00 20.00
HETATM 2 C2 UNK 1 0.002 -0.004 0.002 1.00 20.00
HETATM 3 C3 UNK 1 -0.700 -0.531 1.255 1.00 20.00
HETATM 4 O4 UNK 1 1.020 2.121 0.004 1.00 20.00
HETATM 5 O5 UNK 1 -1.190 2.158 0.019 1.00 20.00
HETATM 6 N6 UNK 1 1.394 -0.475 -0.014 1.00 20.00
HETATM 7 H7 UNK 1 -0.515 -0.370 -0.885 1.00 20.00
CONECT 1 2 4 5
CONECT 2 1 3 6 7
CONECT 3 2
CONECT 4 1
CONECT 5 1
CONECT 6 2
CONECT 7 2
END

46
Crystallographic Information File

 CIF file with internal coordinates

 Parameters for refinement of X-ray structures
<keywords for connectivity>
UNK 'C1' n/a 'C2' START
UNK 'C2' 'C1' 'C3' .
…
UNK 'H7' 'C2' . END
<keywords for bonds>
UNK 'C1' 'C2' single 1.507 0.020
UNK 'C1' 'O4' double 1.208 0.020
…
<keywords for bond angles>
UNK 'C2' 'C1' 'O4' 120.000 3.000
UNK 'C2' 'C1' 'O5' 120.000 3.000
…
<keywords for torsion angles>
UNK var_000 'O4' 'C1' 'C2' 'C3' 120.000 20.0000 6

47
Z Matrix

 Input to quantum mechanical programs

C1 y
y 6 6
C2 1.5 1
Cl3 1.7 1 109 2 z
H4 1.1 2 109 1 -60 3 7 5 r1 5 7
H5 1.1 1 109 2 180 4 1 2 1
x 2 z
Cl6 1.7 2 109 1 60 5 a 4 8 8 4
r2
H7 1.1 1 109 2 -60 6
t4-2-1-3
H8 1.1 2 109 1 180 7 3 3

48
Molecular Surfaces

 Interaction between molecules mediated through surface (properties)

 Electrostatic, hydro/lipophilicity, H-bond donor/acceptor potential

 Common 3D surface models

 van der Waals surface
 Conolly surface
 Solvent-accessible surface (SAS)

 2D approaches
 Topological polar surface area
 Approximate surface area

49
van der Waals Surface

 Simplest representation of molecular surface

 Hard-sphere model
 Determined from van der Waals radii of each atom
 Correction by energetically-favorable distance on non-bonded atoms

Formic acid

van der Waals radius of

van der Waals surface
carboxyl carbon atom 50
Conolly Surface

 Smoother surface (than van der Waals)

 "Rolling" a probe sphere (e.g., solvent molecule) over van der Waals surface
 Water molecule with effective radius of 1.4 Å
 Convex contact surface
 Convex re-entrant surface where probe contacts two or more atom spheres
Contact (Conolly) surface

Re-entrant surface

Formic acid Probe sphere

van der Waals

surface 51
Solvent Accessible Surface

 Similar to Conolly surface, but center of probe sphere defines surface

 Molecular surface that a solvent molecule can access
 Conolly molecular surface extended by effective radius of solvent molecule

Molecular surface
(solvent excluded) Solvent accessible surface

Probe sphere

52
2D Approaches for Surfaces

 Topological polar surface area (Ertl et al., 2000)

 Estimation of polar surface area by sum of contributions of 2D polar fragments
 Contribution derived by fitting of fragment-based TPSA to 3D polar surface area
 Based on 39,000 drug molecules
 43 fragments
 r2 of 0.98 and average error of 5.6 Å
Fragment PSA Frequency
NR3 3.24 0
NHR2 12.03 1
NH2R 26.02 0
TPSA: 49.3 Å
R-O-R 9.23 0
R-O-H 20.23 1
C=O 17.07 1
Query compound
… … …
53
2D Approaches for Surfaces

 Approximate surface area (Labute, 2000)

 van der Waals radii and ideal bond length
 Geometrical considerations to approximate overlap of spheres and remaining
"free" surface
 1,947 molecules
 r2 of 0.96 and average error <10%

54
Molecular Surface Descriptors

 QSAR experiments
 Molecular descriptors independent of size and number of atoms of molecules
 Invariant against translation and rotation of molecules
 Fixed numbers of descriptors for each molecule in data set

 Autocorrelation of surface properties

 Calculation of e.g. electrostatic potential for
points on molecular surface
 Application of autocorrelation formalism
 Sampling of pairwise surface property
points in equidistant intervals 1
A(d n ) = 
2 Ln i , j
pi p j
i j

55
Application – Differentiating Drug
Space
 Dataset of 207 drugs from five different pharmacological actions
 56 ACE inhibitors (ACE)
 49 COX inhibitors (COX) O

 38 folic acid antagonists (FOL) O N

 11 HMG-coA reductase inhibitors (HMG) O N

ACE
 53 monoamine oxidase inhibitors (MAO) O
O
 Most similar COX and MOA O
COX
 Common structural features O O
O,C.N
O O
N N O

N
N,C N,C O
HMG MAO
N N
FOL O Ring N,O
O F Ring
O O
56
Application – Differentiating Drug
Space
 116 CORINA Symphony descriptors
 Global molecular properties
 # H bond acceptors and H bond donors, TPSA, molecular weight, dipole
moment, molecular polarizability, logP and logS
 3D autocorrelation vectors using partial charges, electronegativities and
polarizabilities as atom pair properties
 Autocorrelation of surface properties using molecular electrostatic, hydrogen
bonding and hydrophobicity potential

57
Unsupervised Classification

 Kohonen map trained with CORINA Symphony descriptors

ACE 207 compounds

COX
116 descriptors
FOL
HMG
15x10 neurons
MAO rectangular

 Acceptable separation of classes

 Only 5 conflict neurons
 Only 4 of 10 possible class collisions

58
CORINA Classic – Online Service

 Calculation of 3D structures
 https://www.mn-am.com/online_demos/corina_demo_interactive

CORINA Classic

59
CORINA Symphony CE – Online
Service
 Calculation of molecular properties and descriptors
 https://www.mn-am.com/services/corinasymphonydescriptors

60
Acknowledgements

 BigChem project
 Barbara Gasset, Igor Tetko

 My colleagues at MN-AM
 Aleksandra Mostrag-Szlichtyng, Chihae Yang, Vessela Vitcheva, Aleksey Tarkhov,
Bruno Bienfait, James Rathman, Jörg Marucszyk, Oliver Sacher, Thomas
Kleinöder, Tomasz Magdziarz, Vijay Gombar

 Thank you for your attention!

 www.mn-am.com
61
Selected References

 Databases
 Cambridge Crysatllographic Database
 Allen, F. H.; Kennard, O. Acc. Chem. Res., 1983, 16, 146. c) Allen, F.H.; Hoy,
V.J. Cambridge Structural Database. In Encyclopedia of Computational
Chemistry, Schleyer, P.v.R.; Allinger, N.L.; Clark, T.; Gasteiger, J.; Kollman,
P.A.; Schaefer, III, H.F.; Schreiner, P.R. (Eds.), John Wiley & Sons, Inc.,
Chichester, UK, 1998; pp. 155-167.
 http://www.ccdc.cam.ac.uk/
 The PubChem project
 https://pubchem.ncbi.nlm.nih.gov
 RCSB Protein Data Bank
 http://www.rcsb.org/pdb/home/home.do

62
Selected References (cont.)

 3D structure and conformation generators

 Sadowski, J.; Gasteiger, J.; Klebe, G. Comparison of Automatic Three-
Dimensional Model Builders Using 639 X-Ray Structures. J. Chem. Inf. Comput.
Sci. 1994, 34, 1000-1008.
 Schwab, C.H., Representation of 3D Structures. In Chemoinformatics - A
Textbook, J. Gasteiger and T. Engel, Eds., Wiley-VCH, Weinheim, 2003, 91-124.
 Sadowski, J. Representation of 3D Structures. In Handbook of Chemoinformatics
- From Data to Knowledge, J. Gasteiger, Ed., Vol. 1, Wiley-VCH, Weinheim, 2003,
231-261.
 Schwab, C.H. Conformational Analysis and Searching. In Handbook of
Chemoinformatics - From Data to Knowledge, J. Gasteiger, Ed., Vol. 1, Wiley-
VCH, Weinheim, 2003, 262-301.

63
Selected References (cont.)

 3D structure and conformation generators (cont.)

 Renner, S.; Schwab, C.H.; Schneider, G.; Gasteiger, J. Impact of conformational
flexibility on three-dimensional similarity searching using correlation vectors. J.
Comp. Inf. Model. 2006, 46, 2324-2332.
 Schwab, C.H. Conformations and 3D pharmacophore searching. Drug Discovery
Today: Technologies, Volume 7, Issue 4, Winter 2010, e245-e253.

64
Selected References (cont.)

 Force field calculations

 Leach A.R. Molecular Modelling: Principles and Applications. Prentice Hall, 2nd
Edition, 2001.
 Lanig, H. Molecular Mechanics. In Chemoinformatics - A Textbook, J. Gasteiger
and T. Engel, Eds., Wiley-VCH, Weinheim, 2003, 338-358.
 Lanig, H. Molecular Mechanics. in Handbook of Chemoinformatics - From Data
to Knowledge, J. Gasteiger, Ed., Vol. 3, Wiley-VCH, Weinheim, 2003, 920-946.

65
Selected References (cont.)

 Quantum mechanical calculations

 Leach A.R. Molecular Modelling: Principles and Applications. Prentice Hall, 2nd
Edition, 2001.
 Clark, T. Quantum Mechanics. In Chemoinformatics - A Textbook, J. Gasteiger
and T. Engel, Eds., Wiley-VCH, Weinheim, 2003, 376-400.
 Clark, T. Quantum Mechanics. In Handbook of Chemoinformatics - From Data to
Knowledge, J. Gasteiger, Ed., Vol. 3, Wiley-VCH, Weinheim, 2003, 947-975.

 Structure representation and chemical file formats

 Engel, T. Representation of Chemical Compounds. In Chemoinformatics - A
Textbook, J. Gasteiger and T. Engel, Eds., Wiley-VCH, Weinheim, 2003, 15-90.

66
Selected References (cont.)

 Structure representation and chemical file formats (cont.)

 SD File format
 Dalby, A. et al. Description of Several Chemical Structure File Formats Used
by Computer Programs Developed at Molecular Design Limited. J. Chem. Inf.
Comput. Sci. 1992, 32, 244-255.
 http://accelrys.com/products/informatics/cheminformatics/ctfile-formats/no-fee.php
 PDB file format
 Bernstein, F. C. et al. The Protein Data Bank: A Computer-Based Archival File
for Macromolecular Structures. J. Mol. Biol. 1977, 112, 535-542.
 http://www.wwpdb.org/documentation/file-format
 CIF format
 Hall, S.R. et al. The Crystallographic Information File (CIF): a New Standard
Archive File. Acta Cryst. 1991, A47, 655-685.

67
Selected References (cont.)

 Molecular surfaces
 Engel, T. Representation of Chemical Compounds. In Chemoinformatics - A
Textbook, J. Gasteiger and T. Engel, Eds., Wiley-VCH, Weinheim, 2003, 124-129.
 Connolly M.L. Analytical molecular surface calculation. J. Applied Crystallogr.
1983, 16, 548-558.
 Zhao, Y.H.; Abraham; M.H.; Zissimos, A.M. Determination of McGowan Volumes
for Ions and Correlation with van der Waals Volumes. J. Chem. Inf. Comp. Sci.
2003, 43, 1848-1854.
 Ertl, P; Rohde, B.; Selzer, P. Fast Calculation of Molecular Polar Surface Area as a
Sum of Fragment-Based Contributions and Its Application to the Prediction of
Drug Tansport Properties. J. Med. Chem. 2000, 43, 3714-3717.
 Labute, P. A widely applicable set of descriptors. J. Mol. Graph. Mod. 2000, 18,
464-477.

68
Selected References (cont.)

 Neural networks
 Zupan, J.; Gasteiger, J. Neural Networks in Chemistry and Drug Design. Second
Edition, Wiley-VCH, Weinheim, 1999, 380 pages.
 Zupan, J.; Gasteiger, J. Neural Networks in Chemistry. Angew. Chem. Int. Ed.
Engl. 1993, 32, 503-527.
 Terfloth, L.; Gasteiger, J. Self-organizing Neural Networks in Drug Design.
Screening - Trends in Drug Discovery 2001, 2(4), 49-51.

69