Novel scoring function for modeling structures of oligomers of transmembrane α-helices

Specific non-covalent interactions between transmembrane (TM) α-helices are important in a variety of biological processes. Experimental and computational studies have shown that van der Waals interactions play an important role in the tight packing between TM α-helices, although polar interactions can also be important in some instances. Based on the assumption that van der Waals interaction alone is sufficient for a meso-scale (residue-scale) description of the interaction between TM α-helices, we have designed a novel residue-scale scoring function for modeling structures of oligomers of TM α-helices. We first calculated atomistic van der Waals interaction energies between two amino acids, X and Y, of a pair of parallel α-helices, glycine-X-glycine and glycine-Y-glycine and compiled them according to three variables, the distance between the two C_α atoms and the rotational angles of X and Y about their helical axes. Upon averaging over the rotational angles, we obtained one-dimensional interaction energy profiles that are functions of the distance between C_α atoms only. Each of the interaction energy profiles was fitted with a generic fitting function of the distance between C_α atoms, yielding analytical scoring functions for all possible amino acid pairs. For glycophorin A, neu/erbB-2, and phospholamban, lowest-energy conformations obtained through exhaustive scanning of the entire conformational space using the scoring functions were compatible with available experimental data.