Coevolutionary Landscapes of Kinase Family Proteins

Abstract

The co-variation of pairs of mutations in multiple sequence alignments (MSAs) of protein families can be used to build statistical “Potts” models of the sequence patterns. These models have been found to accurately predict contacts in protein structures, illustrating how evolutionary fitness landscapes of protein families can be connected to corresponding conformational free energy landscapes. We show how a Potts model, parametrized on the protein-kinase family MSA, can further be used to predict the propensity of particular kinase sequences to assume a “DFG-out” conformation, implicated in the susceptibility of some kinases to type-II inhibitors. We also investigate the model's ability to describe kinase mutational statistics. We show how the pairwise (residue-residue) interaction terms of the model are necessary and sufficient to capture the higher-than-pairwise mutation patterns of natural kinase sequences, and can indicate which physical interactions contribute to the fitness and conformational preference of individual sequences, and discuss the number of sequences necessary for model inference.