Towards an understanding of membrane-mediated protein-protein interactions.

We propose a computational framework to study the lipid-mediated clustering of integral membrane proteins. Our method employs a hierarchical approach. The potential of mean force (PMF) of two interacting proteins is computed under a coarse-grained 3-D model that successfully describes the structural properties of reconstituted lipid bilayers of dymiristoylphophatidylcholine (DMPC) molecules. Subsequently, a 2-D model is adopted, where proteins represented as self-avoiding disks interact through the previously computed PMF, which is modified to take into account three body corrections. The aggregation of the proteins is extensively studied under the condition of negative hydrophobic mismatch: the formation of clusters with increasing size agrees with previous computational and experimental findings.