A reagent-based strategy for the design of large combinatorial libraries: a preliminary experimental validation.

Combinatorial library design can be carried out at either the reagent or the product level. Various reports in the literature have come to conflicting conclusions in favor of one over the other. In this paper a reagent-based screening library design strategy is presented. The method relies on analysis of scaffolds and building blocks separately to define the overall diversity in a compound file. The primary diversity selection by properties relevant for molecular recognition and by redundancy is followed by the application of filters for molecular properties known to be relevant for drug-likeness. Filter properties are rapidly estimated at the product level using a fragmental estimation approach. Initial experimental data suggest that high diversity in vast screening libraries can be achieved by carefully applied reagent level analysis. A potential role of diverse screening libraries in chemical genomics (pharmacological knockouts) is also discussed.