skip to content

Yusuf Hamied Department of Chemistry

 

Machine learning promises high-quality predictions at tremendously reduced computational cost compared to standard quantum chemistry methods. Several machine learning models for single-molecule properties have already demonstrated considerable success. However, standard approaches to machine learning in chemistry are not well suited to modeling intermolecular interactions, which govern protein-ligand binding, biomolecular structure, and properties of condensed phases. This talk will explain the challenges for applying machine learning to intermolecular interactions, and our approaches to overcome them. We have examined pure machine learning methods, physics-based models whose parameterization has been accelerated by machine learning, and combinations of the two. The speed and accuracy of the resulting methods will be illustrated for protein-ligand interactions.

[1] Approaches for Machine Learning Intermolecular Interaction Energies and Application to Energy Components From Symmetry Adapted Perturbation Theory, D. P. Metcalf, A. Koutsoukas, S. A. Spronk, B. L. Claus, D. A. Loughney, S. R. Johnson, D. L. Cheney, and C. D. Sherrill, J. Chem. Phys. 152, 074103 (2020) (doi: 10.1063/1.5142636)
[2] AP-Net: An Atomic-Pairwise Neural Network for Smooth and Transferable Interaction Potentials, Z. L. Glick, D. P. Metcalf, A. Koutsoukas, S. A. Spronk, D. L. Cheney, and C. D. Sherrill, J. Chem. Phys. 153, 044112 (2020) (doi: 10.1063/5.0011521)
[3] Electron-Passing Neural Networks for Atomic Charge Prediction in Systems with Arbitrary Molecular Charge, D. P. Metcalf, A. Jiang, S. A. Spronk, D. L. Cheney, and C. D. Sherrill, J. Chem. Inf. Model. 61, 115 (2021) (doi: 10.1021/acs.jcim.0c01071)

Further information

Time:

12May
May 12th 2021
15:00 to 16:00

Venue:

Zoom - link to be announced

Series:

Theory - Chemistry Research Interest Group