Novel algorithms for the optimization of molecular diversity of combinatorial libraries

Various approaches to measuring and optimizing molecular diversity of combinatorial libraries are presented. The need for different diversity metrics for libraries consisting of discrete molecules (cherry picking) vs libraries formed from combinatorial R-group enumeration (array-based selection) is discussed. Ideal requirements for diversity metrics applied to array-based selection are proposed, focusing, in particular, on the concept of incremental diversity, i.e., the change in diversity as redundant or nonredundant molecules are added to a compound collection or combinatorial library. Several distance and cell-based diversity functions are presented and analyzed in terms of their ability to satisfy these requirements. These diversity functions are applied to designing diverse libraries for two test cases, and the performance of the diversity functions is assessed. Issues associated with redundant molecules in the virtual library are discussed and analyzed using one of the test examples. The results are compared to reagent-based diversity optimizations, and it is shown that a product-based diversity protocol can result in significant improvements over a reagent-based scheme based on the diversity obtained for the resulting libraries. (C) 2000 by Elsevier Science Inc.