Predicting water at the protein interface in cryo-EM structures from MD-excess chemical potential

Predicting the positions of water molecules at the protein interface remains a formidable challenge in structural biology, fueling active research in this field. Here, we present a novel approach based on molecular dynamics (MD) simulations that utilize statistical thermodynamic signatures of water at protein interfaces that can be used to improve the accuracy of water placement in maps derived by cryogenic electron microscopy (cryo-EM). We employ an analysis based on the excess chemical potential, or the work to transfer (WT) a water molecule from the bulk to the interface. WT is a measure of the thermodynamic balance between the interaction energy of an interfacial water molecule with the protein and its free energy of interaction with all the other solvent molecules. WT is proportional to the log ratio of the local density of water molecules at the protein interface to the bulk density. Using apoferritin as a benchmark system, we found that 85% of the top 100 water locations with the most favorable excess chemical potential values are observed in one or more structures whose locations were determined from high-resolution cryo-EM maps deposited in the Protein Data Bank (PDB). Seventy percent of the top 200 water locations indexed by excess chemical potential were also observed in PDB structures derived from the cryo-EM maps. The MD simulations are performed without experimental density restraints, yet the water positions with favorable WT values correlate strongly with their corresponding position within experimentally defined maps. This work paves the way for the development of a cryo-EM water placement and refinement tool that integrates MD simulations of the excess chemical potential with cryo-EM data for accurate modeling of water networks.