Generative and reinforcement learning approaches for the automated de novo design of bioactive compounds

oleh: Maria Korshunova, Niles Huang, Stephen Capuzzi, Dmytro S. Radchenko, Olena Savych, Yuriy S. Moroz, Carrow I. Wells, Timothy M. Willson, Alexander Tropsha, Olexandr Isayev

Format: Article
Diterbitkan: Nature Portfolio 2022-10-01

Deskripsi

Deep generative neural networks are increasingly exploited for drug discovery, but often the majority of generated molecules are predicted to be inactive. Here, an optimized protocol for generative models with reinforcement learning is derived and applied to design potent epidermal growth factor inhibitors.