QSAR and Infomatics Tools Case Study

Intertek ASG Computational Chemistry Group

Case Study:

QSAR and Infomatics Tools Applied to Thousands of Real and Virtual Molecules

A major project for one ASG customer has involved an investigation into both the similarity and diversity of a range of different molecules.

This project consisted of two strands: one dealing with real molecules which have been synthesised and one dealing with virtual molecules which only exist in silico.

Product properties such as stability often have a fundamental connection with the structure of key molecular components. Using QSAR techniques a subset of molecules, some of which have experimental stability data was identified.

For each of the molecules, a set of molecular descriptors was generated describing the molecule. Using the descriptors and the experimental data, a model was produced for key chemical parameters which contribute to stability and this has then been applied to try and predict the stability of a molecule for which no experimental data has been measured.

In the second strand a virtual library of over 30,000 structures was enumerated, based on a common fragment with a number of substituent groups.

Descriptors have been generated for each of these molecules and these have been used to categorise the molecules in terms of similarity in molecular space. This has been used to look at the coverage of types of molecules and can also be used in conjunction with the data from the first strand to identify areas of molecular space that give good performance and to identify where molecules have and have not been tested.

An extension to this approach can aid in the selection of molecules to focus synthetic effort on.