Abstract
Water plays a fundamental role in the structure and function of proteins and other biomolecules. The thermodynamic profile of water molecules surrounding a protein is critical for ligand recognition and binding. Therefore, identifying the location and thermodynamic properties of relevant water molecules is important for generating and optimizing lead compounds for affinity and selectivity for a given target. Computational methods have been developed to identify these hydration sites (HS), but are largely limited to simplified models that fail to capture multibody interactions or dynamics-based methods that rely on extensive sampling. Here, we present a method for fast and accurate localization and thermodynamic profiling of HS for protein structures. The method is based on a geometric deep neural network trained on a large, novel data set of explicit water molecular dynamics simulations. We confirm the accuracy and robustness of our model on experimental data and demonstrate its utility on several case studies.