RUI: Collaborative Research: CDS&E: Theory and methods for implicit molecular solvation in ligand and ion binding

The 3D reference interaction site mode (3D-RISM) of solvation is a potentially powerful tool for molecular modeling but its use is hampered by computational cost due to the limitations of current methodologies and CPU hardware. The long-term goal of this research is to create accurate, efficient solvation models for molecular modeling. The objectives of this proposal are to decrease the computational cost of calculating binding affinities of small molecules and ions to large macromolecules by developing new methods for 3D-RISM and utilizing GPU hardware. Our central hypothesis is that 3D-RISM can both accelerate and simplify the modeling of nucleic acids and calculation of absolute binding free energies when coupled with rigorous and efficient solute sampling methods. This was formulated based on our preliminary data and prior researching showing that 3D-RISM can be combined with the double decoupling method (DDM) to rigorously calculate binding free energies, is adept at calculating water and ion density distributions around nucleic acids, and can be accelerated tenfold on GPUs compared to our current implementation on CPUs. Our rationale underlying the proposed work is that it will expedite the screening of diverse drug-candidates for novel lead compounds and study of the structure and thermodynamics of nucleic acids in diverse ionic environments. We are well prepared for this as it builds on recent work we have done to reduce the computation time of 3D-RISM on CPUs, improved the accuracy of the theory, and implement full periodic boundary conditions. We will test our central hypothesis with the following specific aims: 1) combine DDM with 3D-RISM for efficient absolute binding free energies calculations; 2) quantify the effects of ions on nucleic acid structure and thermodynamics; and 3) implement 3D-RISM on GPUs. For the first aim, we will refine our preliminary DDM/3D-RISM protocol to maximize efficiency and create a simple, user-friendly interface. Under the second aim, we will develop and apply robust new 3D-RISM solvers and optimized force fields to predict the solvent environment of nucleic acids over a range of ionic environments. In our third aim, we will implement 3D-RISM on NVIDIA GPUs using the CUDA programming language, which will be distributed with the AmberTools molecular modeling suite as a drop-in replacement for our CPU implementation. Our proposed research is creative and original, in our opinion, as it uses 3D-RISM's strengths to avoid extensive sampling with explicit solvent.