The high throughput screening paradigm of screening as many compounds as possible in a simple low cost screen has largely shifted to screening a focused set of compounds in one or more complex, biologically relevant screens.  In the current paradigm, teams often carry out a subsequent expansion step to identify and screen additional chemical matter, in order to explore structure-activity relationships or to identify additional scaffolds.  We will describe an in-house self-service computational tool, Expansion Master, that enables project teams to drive this expansion step.  It utilizes dozens of orthogonal descriptors (topological, shape-based, activity-based) as well as multiple machine learning methods.  We believe that this tool has set an industry standard in terms of democratizing machine learning methods during lead discovery, by placing multiple complex methods often carried out by different experts, into a single workflow that can be driven and interpreted by non-experts.  It has been applied in more than one hundred projects, and powered both our internal and external facing (FastLabs) screening teams.  The science and architecture behind the tool, as well a comparison of different fingerprints and expansion methods will be discussed.  In addition, we will highlight some specific examples of project impact (scaffold hopping, hit rate enrichments).

AI driven screening with expansion master

ACS Spring 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.