Structural similarity enhances interaction propensity of proteins.

摘要: We study statistical properties of interacting protein-like surfaces and predict two strong, related effects: (i) statistically enhanced self-attraction proteins; (ii) attraction proteins with similar structures. The effects originate in the fact that probability to find a pattern self-match between identical, even randomly organized protein is always higher compared for match different, promiscuous surfaces. This theoretical finding explains prevalence homodimers protein-protein interaction networks reported earlier. Further, our findings are confirmed by analysis curated database complexes showed highly significant overrepresentation dimers formed structurally divergent sequences ("superfamily heterodimers"). suggest homodimeric interactions pose strong competitive heterodimers evolved from homodimers. Such evolutionary bottleneck overcome using negative design pressure applied against homodimer formation. achieved through formation specific contacts charged residues as demonstrated both model real superfamily heterodimers.