Computational prediction of protein–protein binding affinities

TL;DR: The present review aims at presenting an overview on available methods and discussing advantages, approximations, and limitations of the various methods, with a focus on molecular mechanics force‐field approaches.

Abstract: Protein–protein interactions form central elements of almost all cellular processes. Knowledge of the structure of protein–protein complexes but also of the binding affinity is of major importance to understand the biological function of protein–protein interactions. Even weak transient protein–protein interactions can be of functional relevance for the cell during signal transduction or regulation of metabolism. The structure of a growing number of protein–protein complexes has been solved in recent years. Combined with docking approaches or template‐based methods, it is possible to generate structural models of many putative protein–protein complexes or to design new protein–protein interactions. In order to evaluate the functional relevance of putative or predicted protein–protein complexes, realistic binding affinity prediction is of increasing importance. Several computational tools ranging from simple force‐field or knowledge‐based scoring of single protein–protein complexes to ensemble‐based approaches and rigorous binding free energy simulations are available to predict relative and absolute binding affinities of complexes. With a focus on molecular mechanics force‐field approaches the present review aims at presenting an overview on available methods and discussing advantages, approximations, and limitations of the various methods.