(330j) Replica Exchange and Backbone Sampling Methods Improve Protein-Protein Docking By Mimicking Induced-Fit Pathways

Protein-protein interactions (PPIs) are involved in almost all biological processes in human health and disease. Predicting protein complex structures and the associated dynamics of protein interactions can reveal biological mechanisms and suggest intervention strategies. Since investigation of these structural complexes by experimental techniques can be often expensive, laborious and limited, computational modeling provides an alternative. Current computational methods are confounded by the many degrees of freedom of the binding-induced conformational changes in one or both protein partners. To overcome these limitations in capturing backbone motions, we developed a new, aggressive conformational sampling method that incorporates temperature and Hamiltonian replica exchange Monte Carlo (T-REMC and H-REMC) techniques within docking protocols in Rosetta. To mimic the induced-fit approach of protein binding, we sample backbone motions on residues comprising a putative interface on-the-fly, thereby enabling the algorithm to capture protein backbone flexibility while docking. Our updated method ReplicaDock 2.0 requires 300-500 CPU hours, depending on protein sizes, which is efficient compared to molecular dynamics-based docking approaches. In local docking of a benchmark set of 80 proteins with moderate to high flexibility (unbound to bound RMSD over 1.2 Å), ReplicaDock 2.0 achieves successful docking for 61% of moderately flexible complexes and 33% of highly flexible complexes. Additionally, highly flexible targets can be predicted to sub-angstrom accuracy when backbone sampling is biased towards residues comprising flexible loops or hinge domains, indicating that additional gains are possible when mobile protein segments are known.