About the Molecules

News 2 May 2001

We hope to start using revised configurations this week with approximately 35 million drug-like molecules being downloaded plus 100 derivatives of each (i.e. 3.5 billion molecules in total). This will use a revised version of the software which will address a number of the concerns raised in feedback to date and analysing the results. These will mostly be at the minimum conformer precision, 200 molecules per work unit and 100 derivatives per work unit. After reviewing any work units which take abnormally long, we plan to repeat these at a higher conformer precision (probably with the new faster THINK v1.11 or v1.12 - more info in the news section). By doing so, we hope to learn a lot about improving the methodology as well as finding additional interesting molecules.

Have all the molecules being processed previously been made?

No. The 2.5 million starting molecules include those molecules which are commercially available and are suitable – but this is only about 0.4 million. A further 2.1 million suitable molecules originate from approximately 20 well-defined Combinatorial Chemistry libraries.

How do decide what is a suitable molecule?

We have attempted to eliminate molecules that do not have “drug-like” physical properties including those that are unlikely to be soluble, are reactive or easily metabolized etc. About 6 million molecules have been preprocessed in order to generate the 2.5 million starting molecules.

How many molecules are processed in each work unit?

100 molecules are downloaded and the software attempts to generate 100 suitable derivatives are made by small changes to these molecules giving a maximum of 10,100 per work unit. Thus 2.5 million starting molecules mean that we process 250 million molecules

Are all the molecules unique?

When processing commercial catalogues of molecules we have eliminated duplicates but it is still possible that some of the derivatives from different molecules will be the same. The scientific publications of this project will include a statistical prediction of the overlap.

Molecular Catalogue Donors Several organizations generously donated their collection of chemical data, or "catalogues" of molecules to contribute to this project. These contributors include: Asinex; ChemBridge, ComGenex, Contact Services (now known in the USA as ChemDiv Inc.), Maybridge , and Oxford Molecular's HTS Database (sourced from Bionet, ChemStar, Labotest, Orion, Sigma-Aldrich, Specs & Biospecs and TRG)

Figure. An example combinatorial molecule - not a commercially available chemical but if it proved to have characteristics which interact with the target protein it will be investigated further.