Deep learning-based approaches are increasingly useful for improving aspects of small molecule lead optimization for drug discovery. Advancements have been fueled by the generation of large datasets generated in a standardized manner, increasing computing power, and the adaptation of algorithmic advances to chemistry. Here, we highlight learnings from our application of deep learning methods to lead optimization problems. We first show how graph neural networks result in significant prediction performance advances for several key small molecule ADME properties, and we explore the basis of that improvement. We then show how the learned QM potential, ANI, from Roitberg and co-workers can contribute to improvements in small molecule lead optimization. We finally conclude with our recent work marrying the two approaches, which shows promise and a new hope.

Deep learning for medicinal chemistry lead optimization: Dawn of a new hope

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.