To improve computational modeling in biology, we need to deepen our understanding of water and improve our models of solvation. Explicit water models are computationally expensive and implicit water models miss much of the physics, so computer simulations of biomolecules often don't predict experiments as well as they could. We propose here a new approach to solvation that aims to be as accurate as explicit models and as fast as implicit models. We have three aims: (1) To develop 3D analytical and integral-equation approaches to compute structures and energetics of water, (2) To compare explicit with implicit solvation simulations to learn the nature of water structuring in solvation shells, and (3) To develop a Semi-Explicit method for solvation, which is faster than explicit, and more physical than implicit. Our approach is based more on the local statistical mechanics of each water molecule, rather than on continuum approximations (implicit), or brute force stochastic simulations. Our preliminary results give us optimism that this approach is working. Our model gives the density of water vs. temperature as accurately as TIP4P-Ew but 6 orders of magnitude faster. The preliminary phase diagram of water looks good. Our solvation model is capturing the free energies of solvation of neutrals and polar solutes about as accurately as explicit, and is about as fast to compute as GB. Our recent results in the blind SAMPL computational solvation modeling event are highly encouraging.
The foundation of biological processes starts at the molecular level, and one of our key tools for understanding microscopic systems is computational modeling. Computer simulations of biomolecules often don't predict experiments as well as they could, and one of the primary reasons is limitations in the modeling of ever present water. We propose to develop new approaches for treating water that aim to deepen our understanding and lift the limitations of models for solvation.
Showing the most recent 10 out of 74 publications
