Computational characterization of SAR microenvironments in high-throughput screening data

Abstract

A computational approach is described to analyze structure-activity relationship (SAR) information contained in compound and screening data sets. The methodology is designed to explore SAR information in a systematic and compound-centric manner in order to aid in the selection of hits from high-throughput screening (HTS) data. Methods: Chemical neighborhood graphs integrate a graphical representation of the chemical environment of each active compound in a data set with the potency distribution within its neighborhood and information from a quantitative SAR analysis function. Environments are systematically generated and ranked by SAR information content. From these environments, key compounds and compound series can be selected. Results: The methodology is described in detail. In addition, the application to four screening data sets is reported, revealing different SAR characteristics. A number of different examples of compound environments are presented and discussed that have varying SAR information content. Conclusion: Chemical neighborhood graphs provide an intuitive graphical access to SAR information contained in hit sets. SAR information is analyzed in a compound-centric manner, with a focus on local SAR environments (microenvironments). It is anticipated that this approach will complement and help to further refine current hit selection strategies and trigger the development of additional graphical analysis methods to search for SAR information in HTS data.