This work has proceeded along several fronts.  I have developed ML models for the prediction of protein-ligand binding and applied those models to systems underlying plant-microbe symbiosis and SARS-Cov-2 virulence.  To address the problem of how binding modulates the protein function, I am also working on using convolutional neural networks to predict the allosteric state of a protein.  Along the same theme of predicting molecular interactions and their functional consequences, colleagues and I in the Biofuels SFA are using ML to predict interactions of lignocellulosic biomass with organic solvents with the ultimate goal of predicting which solvents deconstruct biomass most efficiently so that lignin, cellulose, and hemicellulose can be valorized as part of a green economy with a low carbon footprint.

I have also developed a technique where experimental small-angle scattering is used to tune the approximate potential energy models that are used in molecular dynamics.  This has been applied to intrinsically disordered proteins and multicomponent lipid membranes.

Lastly and most recently, I have applied a combination of ML and physics to problems in enzyme engineering, including predicting product yields of diverse panels of putative enzymes and predicting the effects of site mutagenesis on enzyme function.  The ultimate goal of this work is to break down and upcycle plastics using engineered enzymes, as plastics have become a significant pollutant owing to their inability to be recycled by the current state of the art.