Artificial intelligence (AI) is being increasingly applied in drug discovery to make the process more efficient and less costly. In this talk we present one of the first applications of REINVENT(J. Chem. Inf. Model. 2020, 60, 12, 5918–5922), an AI-based de novo design tool, on an active project involving a kinase target. The goal of the effort is to identify a new chemical series to supplement the lead series identified through high-throughput screening. The chemical space was defined by the training set, which included over 1 million molecules representing compounds from ChEMBL and AstraZeneca. The reinforcement learning was guided by a multi-parameter optimization (MPO) function, which included the activity against main target, the selectivity against key off-targets, and PK properties. Historical potency and ADME data were used to create QSAR models for the MPO function. The output molecules from REINVENT were selected by docking and in-house in silico ADME models. Ideas inspired by REINVENT output were also added into the final list. Three different chemical series were identified by this effort. Scaffold A represented a new kinase scaffold. The synthesis was challenging, however the compound made demonstrated ~100nM potency against the target. Scaffold B was a known kinase scaffold, but all previous examples with this scaffold were not active against this kinase. REINVENT generated new compounds with single-digit nanomolar potency. Since this scaffold was amenable to library-synthesis, we were able to explore SAR very quickly. Scaffold C had been identified by HTS previously, but had not been actively pursued. REINVENT-inspired ideas led to improved selectivity for this series. The three series exemplifies the potential applications of REINVENT in both hit finding and hit-to-lead optimization. We will also discuss the lessons learned through this process.

Application of REINVENT, an AI-based de novo design tool, in real-life drug discovery

Fall 2022 abstracts, posters, and presentations will launch in mid-September.

ACS Fall 2022

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