(695a) Predictive Enzyme Catalysis and Protein Structure with Quantum Chemistry On GPUs

Proteins are large macromolecules that play a major role in nearly every basic function that occurs in the cell, from enzymes that speed up reactions within the cell to membrane proteins that regulate the cellular environment. Miscues in protein folding or function are often responsible for the onset of diseases, and an understanding of underlying structure-function relationships in these proteins is critical for disease treatment. Nevertheless, computational study of proteins is often restricted either to empirical, force-field approaches on large systems or highly-accurate approaches on small model systems. Massively parallel graphical processing units (GPUs) permit speed-ups for computational chemistry approaches.  Recent tailoring of electronic structure methods for GPUs enables us to apply quantum-mechanical methods to proteins as large as several thousand atoms. We will highlight the increased understanding that our GPU-accelerated, quantum-mechanical approach has provided in the enzyme catechol-O-methyltransferase and the light-driven proton pump proteorhodopsin.  In particular, our work demonstrates that much larger quantum mechanical regions in commonly employed QM/MM methods may be needed in order to achieve predictive accuracy for theoretical descriptions of enzyme catalysis mechanisms. We will also show how interfacing with advanced density functional theory methods allow us to reach predictive accuracy in descriptions of key properties such as protonation states in proteins.