A deep-learning model for grid-based solvation free energy.

Abstract

Listen

Translate article

In computational drug discovery, the stability of protein-ligand binding is estimated by calculating the free energy change before and after binding. However, one of the factors of the free energy change is the free energy change of water molecules around the protein (hydration free energy), which requires a huge computational cost. We focus on deep learning and aim to develop a deep learning model that can calculate the hydration free energy at high speed. In this presentation, we show the results of constructing a deep learning model based on hydration free energy data obtained by grid inhomogeneous solvation theory (GIST) and evaluating its prediction performance.