• Lipinski rule-of-5, Veber criteria and reactive and undesirable chemical group filtering
• In silico docking of physically available and virtual compound collections into enzyme and receptor active sites
• Two- and three-dimensional similarity and pharmacophore searching for molecules expected to have the same binding characteristics as known lead molecules
• Chemical complementarity of what has already been tested by the customer or what is known from literature

BioFocus DPI takes a pragmatic approach to compound selection. Where appropriate we will combine complementary approaches to create smart screening sets. Assembled molecules identified by in silico methods (perhaps just a few hundred molecules) are subsequently screened. Hits can then be expanded into larger sets by searching for similar molecules to those discovered by the biological screening applying two-dimensional (2D) and three-dimensional (3D) similarity approaches.

Benefits

Part of our comprehensive suite of hit-finding technologies, intelligent hit finding is a powerful method of reducing the number of compounds to be physically screened by creating an intelligent subset of molecules based upon knowledge of the structure of the target protein or of its ligands. This iterative testing of subsets of molecules can greatly reduce the cost for the hit-finding process for suitable projects.

Technologies

Enrichment of compound screening sets

Kernel Discriminant Analysis: Public-domain algorithm used to generate screening sets are i) free from undesirable groups and ii) which contain molecules which have been partitioned into desired chemistry spaces.

Thematic Analysis and kinase 2D Roadmap: Proprietary technologies used to give screening sets which are enriched with molecules active against proteins within GPCR and kinase target classes.

2D and 3D similarity searching for ligand-based approaches to hit finding

Recursive Binary Trees: Combination of commercially available and proprietary approaches allowing the identification of related molecules to known or identified hits based on various 2D criteria. Activity and selectivity data can be included. Various sets of 2D molecular descriptors are used derived from commercial and proprietary tools available at BioFocus DPI.

Prediction of activities: Prediction of Activity Spectra of Substances (PASS) database and known compounds contained in databases like MDDR, CMC, IDDB and SciFinder are used to identify similar compounds based on 2D searching criteria.

3D pharmacophore searching: Combination of commercially available tools used to derive 3D pharmacophore hypotheses and perform 3D pharmacophore searches among physically available and virtual compound collections.

Virtual docking toolkits for protein structure-based approaches to hit finding

Industry standard docking software + Refinery: Combination of industry-standard software and Refinery, our proprietary database platform for running and analyzing high-throughput virtual docking of ligands into protein active sites. This platform can be tuned to optimize hit enrichment for any required target for which high-quality protein structure data is available.

Compound databases used for compound searching

Databases used for virtual screening and 2D/3D similarity searching: physically available screening collection at BioFocus DPI (700 K compounds), BioFocus DPI virtual library (1.6 M compounds) and additional third party knowledge bases and commercially-available compounds (3.6 M compounds)

Case studies

1. Virtual screen for kinase inhibitors

In a recent collaboration with a client, we were asked to virtually screen 1.6 M commercially available compounds to identify novel kinase inhibitors. Using the KDA technology, compounds were ranked according to their ‘chemical distance’ to the training set. As a result of this, 2,200 compounds were selected for in vitro screening by the client and subsequently novel hit structures were identified.

2. Structure-based compound selection

Having performed HTS with more than one million compounds from their in-house compound collection for a protease target, the client aimed to find additional hits from a complementary compound collection suitable as back-up candidates for their IND candidate compound.

Given that a number of X-ray structures of the required target in complex with inhibitors exists, a  smart screening campaign was selected using a structure-based compound selection approach. The aim was to identify the maximum number of potential hits from the BioFocus DPI compound library by screening a limited number of compounds.

An enrichment factor of 9.4 compared to random screening could be achieved. After final profiling, 12 compound classes with IC50 values from < 20 μM to < 1 μM could be identified, five of which were of particular interest to the customer. The majority of these compound classes had been identified by the virtual screening protocol.