(346w) Generating Potential Energy Landscapes for Membrane Proteins Using High-Throughput Simulations

Integral membrane proteins are ubiquitous in biological cellular and subcellular membranes. Despite their significance to cell function, isolation of membrane proteins from their hydrophobic lipid environment and further characterization remains a challenge. Several computational approaches, such as docking or self-assembly simulations, have been used to obtain insights into membrane proteins; however, the promise of these approaches has been limited due to the computational cost. Here we present a new method called Protein AssociatioN Energy Landscape (PANEL) that provides an extensive and converged data set for all possible conformations of membrane protein associations using a combination of stochastic sampling and equilibration simulations. The PANEL method generates a potential energy landscape describing the thermodynamic favorability of protein-protein association states, and a frequency landscape representing their kinetic ease of formation. Based on this comprehensive dataset, we constructed a systematic approach to rank the dimer orientation taking both relative stability of the protein dimers and their formation frequency into consideration. We mapped this data to the features on the landscape—Funnels—deep thermodynamically favored states that may be kinetically less favored, and—Saucers—shallow, thermodynamically less stable states, yet wide landscapes that are kinetically favored. Using the Funnel and Saucer characterization of the landscape, we demonstrate an efficient way of identifying significant protein-protein interactions from a diverse set of protein families. The PANEL method is applicable to all membrane proteins and is freely distributed for use by the research community.