Functional group fingerprints: Augmenting hit and lead identification

CINF 82

James R. Arnold, Charles L. Lerman, and James R. Damewood. CNS Chemistry, AstraZeneca, 1800 Concord Pike, Wilmington, DE 19850
It has been estimated that roughly 70% of drug discovery projects must be approached by ligand-based methods, as many targets are not currently amenable to structural studies. We present a novel ligand-based method called Functional Group Fingerprinting. This method classifies medicinally relevant functional groups in molecules using approximately 400 defined functional groups. It creates bitstrings that are used to calculate similarity scores between known actives and either databases or libraries of compounds. In this presentation we will show completeness and orthogonality of the functional group assignments in medicinally relevant compounds. Functional Group Fingerprinting also recovers different sets of actives than those recovered with other fingerprint methods. We will show the enrichment rates observed with this method in greater than 500 target classes within the MDDR are comparable or superior to existing methods. The method recovers on average greater than 60% of the active compounds in less than 1% of the ranked MDDR target classes. This permits fewer compounds to be screened in the hit or lead identification stages in order to identify a sufficient number of chemical classes for lead generation to progress on a given target. That results in reduced depletion of the corporate compound collection, more targets being evaluated, and more efficient identification of lead series for prioritization.