Thomas E. Hadfield

This paper evaluates machine learning-based virtual screening models' ability to identify functional groups responsible for binding using synthetic data and finds that deep learning models outperform traditional methods.

The paper develops a machine learning scoring function for virtual screening that avoids dataset biases and demonstrates its ability to identify important binding interactions through input attribution.

STRIFE improves fragment elaboration by incorporating pharmacophoric information from protein targets, outperforming existing methods in generating ligand-efficient structures.

This paper introduces DEVELOP, a generative model that incorporates 3D pharmacophoric constraints to improve molecule design quality and control.