Dive Deep into Automated Drug Discovery: How Scalable Generative Platform Discovers New Lead-Like Molecules in Days
, Inslico Medicine
, Insilico Medicine
We describe a scalable distributed platform — Chemistry42 — that can be deployed in the cloud and on-premises environments with predictable hardware-agnostic workload management. The platform operates beyond existing screening libraries and skips the effort of scaffold search and structure optimization. Chemistry42 is a fully-automated machine learning platform that delivers new lead-like structures in a week. The platform utilizes NVIDIA GPUs to train and evaluate machine learning models, simultaneously run multiple simulations, and perform quick similarity searches against large databases. We run Chemistry42 on a Kubernetes cluster; the cloud infrastructure automatically manages AWS instances to enable extra NVIDIA GPU nodes when necessary. A presented generative platform is currently being deployed to multiple big pharmaceutical companies and has also been used internally to produce preclinical candidate compounds.