Next generation implicit solvation for atomistic modeling

Project Summary. This proposal responds to PAR-19-253 “Focused Technology Research and Development”. Our main goal is to develop a novel class of implicit solvation models, as accurate, and even more accurate, than standard explicit solvent models, but much faster. The high accuracy, fast implicit solvation models will be combined with several innovative strategies to deliver new computational protocols to improve accuracy and speed of binding free energies prediction, directly relevant to drug design. We will develop a computational tool for fast screening of existing and potential multiple simultaneous mutations in the SARS-CoV-2 coronavirus genome for high affinity to human cells, which translates into high infectivity. Progress in modern bio-molecular sciences, from structural biology to structure-based drug design, is greatly accelerated by atomic-level modeling and simulations that bridge the gap between theory and experiment. The so-called implicit solvation models can provide critical advantages in speed and versatility through representing the effects of solvent – often the most computationally expensive part of such simulations – in a particularly efficient manner. The resulting speed-up of modeling efforts is critical in many areas such as protein folding or protein-ligand docking; however, the accuracy of the current fast models does not reach the standard of the more traditional, but computationally very demanding explicit solvent approach. As a result, prediction reliability of the practical, fast implicit solvation models remains low. In general, high accuracy is a prerequisite for quantitative in-silico drug design. Here, the accuracy limitation of the current implicit solvation framework will be addressed in a novel, systematic way; advantages of the new implicit solvation models will be demonstrated in the context of improving the accuracy of protein-ligand binding free energy calculations. We will use a novel approach to systematically add most of the missing explicit solvation effects to the very basic, but computationally efficient implicit solvation framework of the Poisson and generalized Born (GB) models, with little computational overhead. The GB model is particularly well suited for molecular dynamics simulations. We have set high accuracy standards for the new theory: one kT (thermal noise) deviation from experiment for small molecules hydration, which is better than what most widely used explicit water models, such as TIP3P, can currently deliver. Based on preliminary results, this goal is within reach. The high accuracy combined with the expected computational efficiency will usher in the next generation of implicit solvation models that can make a profound difference in bio-medically relevant atomistic calculations. Example of an immediate impact: Close to, or better than, “industry standard” accuracy in protein- ligand binding calculations, but at a significantly reduced computational expense. Example of a long term impact: Fast, versatile computational tools to analyze binding of viruses to host cells will increase our preparedness for future pandemics.