Machine learning has been successfully applied to various drug-design challenges, including binding-affinity prediction, virtual screening, and QSAR. We here present DeepFrag, a deep convolutional neural network for lead optimization. Given the structure of a protein/ligand complex and the location of a ligand atom, DeepFrag predicts the structural fingerprint of an optimizing moiety that might be added to the ligand to improve binding. It then searches a database of known moieties to find those with similar fingerprints. Using a benchmark test set of known ligands with missing (deleted) fragments, we showed that DeepFrag selects the missing fragment from a set of over 6,500 about 58% of the time. Even when it selects the “incorrect” fragment, the top fragment is often chemically similar and may represent a valid substitution. DeepFrag is freely available for download. We have also implemented it as a Google Colab and stand-alone browser application. This talk will focus on recent DeepFrag developments, applications, and use.

DeepFrag: Deep learning applied to structure-based lead optimization

Where thousands of chemistry professionals meet to share ideas and advance scientific and technical knowledge. Spring 2023 abstracts, posters, and presentations will launch at the end of March.

ACS Spring 2023

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