David Sherrill is supported by an award from the Chemical Theory, Models and Computational Methods program in the Chemistry Division and the Computational Science and Data-enabled Science and Engineering Program to develop computational methods, algorithms and software for studying non-covalent interactions. Non-covalent interactions are central to protein folding, drug docking, crystal packing, and solvation. Recent experiments provide detailed data on organic model systems meant to isolate and quantify individual non-covalent interactions in solution, yet they remain hard to understand without the aid of theory. This project seeks to unravel the various factors influencing non-covalent interactions in these experimental organic systems. The Sherrill group is (1) Developing faster algorithms for density functional theory (DFT) geometry optimization and frequency analysis, using density fitting techniques. (2) Expanding the available set of benchmark-quality non-covalent interaction energies from a few hundred to several thousand, using test cases taken directly from the Protein Data Bank (PDB). This data is critical for validating and parameterizing theoretical models of non-covalent interactions. (3) Identifying synergistic approximations in coupled-cluster theory to accelerate the computations in Goal (2), using combinations of density-fitting, natural orbitals, and rank-reduction techniques. (4) Performing detailed studies of experimental model complexes of CH/pi and pi-pi interactions to better understand these interactions, how they operate in solution, and how they might be more effectively modeled.
The Sherrill group develops and uses computational methods to better understand non-bonded interactions between molecules. Such interactions control how drugs find their targets and how molecular crystals form. Hence, this research provides knowledge useful for advances in rational drug design and crystal engineering. This project provides improved theoretical and computational methods for modeling such non-bonded interactions through (a) improved computer algorithms to compute interactions more quickly, and (b) generation of a large database of high-quality benchmark data that can be used to test and improve new theoretical models of non-bonded contacts. The benchmark data will be made publicly available through a web portal. The software developed will be freely released under an open-source license. This project provides training for students not only in how to use molecular modeling software, but also in how to write molecular modeling software.
