Landscapes of Membrane Protein Interactions from High-Throughput MD Simulations using the Daft Approach

Abstract

Interactions between membrane proteins are key in many biological and pathological processes and offer potential targets for pharmacological intervention. Unfortunately, the complex environment makes it difficult to explore these in high detail. In addition, the time scales of binding and unbinding pose difficulties for molecular simulations to probe such interactions. Over the past few years, we have developed an approach using large numbers of simulations, which avoids the problem of unbinding, allowing rapid building of a detailed map of the interaction landscape. The method, called Docking Assay For Transmembrane components (DAFT), has to date been used to investigate a range of 23 Glycophorin A mutants, a set of 40 receptor tyrosine kinase (RTK) pairs, SNARE protein TM helices, the DesK minimal thermosensor and GPCRs, accounting for >15M CPU hours and representing a total simulation time of more than 30 milliseconds. The results show that several hundreds of simulations are necessary for a converged view and that the time scales required range from 300 ns per simulation for simple helices to microseconds for larger and more complex systems. Yet the results also provide unique views on the convergence properties of ensembles of simulations, yield detailed maps of interaction landscapes, and allow 2D-PMFs to be derived. Furthermore, the comparison of different members of the RTK family of or wild type proteins and mutants gives insight in the mechanisms underlying the relative propensities to dimerize.