While the vaccination against SARS-CoV-2 virus is under way worldwide, therapeutic interventions are still impactful to those patients affected by Covid-19. Efforts to discover inhibitors of the main viral protease 3C-like protease (3CLpro) have been underway since its sequence became available over a year ago. The Moonshot initiative by the PostEra group has led the collaboration of scientists around the world in discovering 3CLpro inhibitors by synthesizing and testing proposed inhibitory molecules. Previously, we used Cyclica’s deep-learning Ligand Design technology supplemented with small fragments identified by the XChem facility (UK) to propose 6 ligands for PostEra’s initial challenge in March 2020, with one molecule possessing measurable activity deemed insufficient for followup. In the current study, we have revisited our initial ligand design experiments, and supplemented our predictions by utilizing the results of the ongoing in vitro testing of molecules submitted to the Moonshot initiative. As of 2021 January 15th, 1477 molecules have been synthesized and tested by PostEra. In this talk, we will describe how we used the activity measurements from those assays to build an activity prediction model using our POEM technology (Pareto-Optimal Embedded Model), a supervised machine learning algorithm based on chemical similarity. Following that, we describe Ligand DesignTM (LD), Cyclica’s platform that traverses a chemical space, either iteratively or exhaustively, and derives molecules that best satisfy a defined objective function, to design candidate inhibitors. We will then describe how we utilized two objective functions (3CLpro binding site using Cyclica’s proteome-screening engine MatchMakerTM, and the POEM model built using the Moonshot data described above) to guide compound exploration and selection for downstream testing.

Machine learning-driven design of candidate inhibitors for the 3CL protease of SARS-CoV-2

ACS Spring Meeting 2021

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