Abstract
Among the several goals of a high-throughput screening campaign is the identification of as many active chemotypes as possible for further evaluation. Often, however, the number of concentration response curves (e.g., IC50s or Kis) that can be collected following a primary screen is limited by practical constraints such as protein supply, screening workload, and so forth. One possible approach to this dilemma is to cluster the hits from the primary screen and sample only a few compounds from each cluster. This introduces the question as to how many compounds must be selected from a cluster to ensure that an active compound is identified, if it exists at all. This article seeks to address this question using a Monte Carlo simulation in which the dependence of the success of sampling is directly linked to screening data variability. Furthermore, the authors demonstrate that the use of replicated compounds in the screening collection can easily assess this variability and provide a priori guidance to the screener and chemist as to the extent of sampling required to maximize chemotype identification during the triage process. The individual steps of the Monte Carlo simulation provide insight into the correspondence between the percentage inhibition and eventual IC50 curves.