• http://ich.vscht.cz/~svozil/teaching.html

EXPLORING CHEMICAL SPACE FOR DRUG DISCOVERYDaniel Svozil svozild@vscht.cz

Laboratory of Informatics and Chemistry

University of Chemistry and Technology Prague

What is chemical space?

size: 3 × 1023 stars, 1080 atoms Dokkum & Conroy, Nature, 2010, 468, 940–942

http://www.universetoday.com/21528/sweet-galactic-molecule-could-point-to-alien-life/

Size of chemical space• mono- to 14-substitute n-hexanes … 1029

Weininger, In Encyclopedia of Computational Chemistry, 1998, Vol. 1, 425-530

• estimates vary wildly, commonly given … 1060 (MW<500, stable, not all synthetically available)Bohacek et al., Med. Res. Rev., 1996, 16, 3-50

• CAS … 1.0 × 108

• ZINC … 2.3 × 107

• DrugBank … 7 759 drugs

all numbers as of 1. 9. 2015

Why we need to explore chemical space

Lipinski & Hopkins, Nature. 2004, 432, 855-61

chemical space

gene family

administered drugs

Methods of its exploration• experimental

• synthesis – combinatorial chemistry• related biological data – high-throughput screening (HTS)

• computational

http://www.cam-com.com/images/comchem.jpghttp://chemgen.cz/

Computational exploration of chemical space

This is basically de novo design. It means that new chemotypes with desired effects are proposed.

Two major approaches

1. Exhaustive enumeration

2. Molecular evolution

Once you have a virtual library generated, you can apply any of possible virtual screening methods to prioritize your compounds.

Exhaustive enumeration• prof. Reymond, Bern• GDB databases, all molecules that can exist up to a

certain number of heavy atoms• GDB-11 ( compounds with C, N, O, F)• GDB-13 ( compounds with C, N, O, S, Cl)• GDB-17 ( compounds with C, N, O, S, halogens)

Molecular evolution• systematicaly explore chemical space by clever generation of new

compounds

Kawai et al., J. Chem. Inf. Model. 2014, 54, 49-56.

Virtual screening ‘funnel’

~106 – 109

molecules

VIRTUAL SCREENING

INACTIVES

HITS

GENERATED DATABASE (Q)SAR

Docking

Pharmacophore models

~101 – 103

molecules

From the presentation by A. Varnek, University of Strasbourg

Filters

Molpher• Molecular morphing

Svozil et al., J. Cheminform., 2014 , 6, 7.

Morphing between two molecules

Svozil et al., J. Cheminform., 2014 , 6, 7.

Morphing operators

Svozil et al., J. Cheminform., 2014 , 6, 7.

Molpher – what is it good for?• Idea: start and target molecules are active against the

same target• In a systematic way, generate chemical subspace

between a start/target pair• Explore this subspace for molecules with high potency

IMG library

ChEMBL

GR ligands

Data cleaning and consolidation Molpher

Pool of posible ligands

ZINC

Purchase compounds

Experimental verification

297 GR ligands87 912 paths

~ 2 000 000 morphs

ACKNOWLEDGMENTLICH group: Ctibor Škuta, Milan Voršilák, Ivan Čmelo,

Martin Šícho, Jiří Novotný

IMG group of chemical biology: Petr Bartůněk, David Sedlák and others