Short Courses in Chemoinformatics

from 17th May 2010 to 21th May 2010

Université de Strasbourg, Faculté de Chimie, Strasbourg, FRANCE

Reference: CBT09-0684A

Program

Day 1

Morning

Historical overview

Computer representation of chemical structures

Software: ChemAxon Marwin

Content: 1D, 2D, 3D and 4D presentation of structures. Adjacency and distance matrices, connectivity tables. Linear notations SMILES, SMARTS, INChI, exchange formats MOL, RXN, SDF and RDF, PDB. Bitsctrings: structural keys and fingerprints, bits collision.

Afternoon

Creation and management of chemical databases

Software: Chemaxon InstantJchem, MOE

Content: Presentation of most important public databases, creation of a chemical database from the scratch, manipulations of data (search, import/export), data curation, databases fusion.

Day 2

Morning

Descriptors

Software: MOE, CODESSA PRO

Content: Generalities, molecular descriptors, concepts of 0D, 1D, 2D, 3D and 4D descriptors, detailed presentation of some frequently used descriptors.

Afternoon

Conformational sampling

Software: MOE

Content: Force Field approach, potential energy surface, systematical and stochastic approaches for conformational sampling.

Day 3

Morning

Pharmacophores

Software: MOE, LigandScout

Content: Intermolecular interactions and pharmacophore concept, 2D and 3D pharmacophores, pharmacophore editing, structure-pharmacophore match, pharmacophore elucidation, hypothesis generation

Afternoon

Chemical space, similarity/diversity and chemical library design

Software: Chemaxon Synthesizer, MOE

Content: Notion of chemical space, similarity/diversity approach, metric (Euclidean, Tanimoto, Manhatan, Max), chemical library design (MinMax, MaxSum, cherry picking, clustering), combinatorial libraries (generation of compound database from a reactant database), focused libraries.

Day 4

Morning and afternoon

Machine Learning

Software: Weka, ISIDA

Content: QSAR/QSPR, short historical overview. Regression and classification models. Statistical parameters. Workflow of obtaining and validation the models. Variables selection: stepwise and genetic algorithms. Classification: Naïve Bayes, Decision Trees, SVM. Regression: linear (MLR, PLS) and non-linear (Neural Networks, SVM). Model validation: cross-validation, y-randomization, bootstrapping. Models Applicability Domain approaches: bounding box, z-kNN, etc.

Day 5

Morning

Docking

Software: MOE

Content: The docking paradigm (similarity of binding), conformer search, pose evaluation, docking score

Afternoon

Virtual screening

Software: ISIDA, MOE

Content: Approaches used in VS: filters,similarity/pharmacophore search, QSAR/QSPR models, docking. Choice of "reasonable" models (descriptors, mathematical relations, applicability domain), ensemble strategies, risks/cost evaluations.