Interfaces Between Alpha-helical Integral Membrane Proteins: Characterization, Prediction, and Docking

Abstract

Protein-protein interaction (PPI) is an essential mechanism by which proteins perform their biological functions. For globular proteins, the molecular characteristics of such interactions have been well analyzed, and many computational tools are available for predicting PPI sites and constructing structural models of the complex. In contrast, little is known about the molecular features of the interaction between integral membrane proteins (IMPs) and few methods exist for constructing structural models of their complexes. Here, we analyze the interfaces from a non-redundant set of complexes of α-helical IMPs whose structures have been determined to a high resolution. We find that the interface is not significantly different from the rest of the surface in terms of average hydrophobicity. However, the interface is significantly better conserved and, on average, inter-subunit contacting residue pairs correlate more strongly than non-contacting pairs, especially in obligate complexes. We also develop a neural network-based method, with an area under the receiver operating characteristic curve of 0.75 and a Pearson correlation coefficient of 0.70, for predicting interface residues and their weighted contact numbers (WCNs). We further show that predicted interface residues and their WCNs can be used as restraints to reconstruct the structure α-helical IMP dimers through docking for fourteen out of a benchmark set of sixteen complexes. The RMSD100 values of the best-docked ligand subunit to its native structure are <2.5 Å for these fourteen cases. The structural analysis conducted in this work provides molecular details about the interface between α-helical IMPs and the WCN restraints represent an efficient means to score α-helical IMP docking candidates.

MENSAdb: a thorough structural analysis of membrane protein dimers

“…The highest differences were for the GAS residues of the core region (core: 0.06 ± 0.03, surface: 0.03 ± 0.02; interface: 0.03 ± 0.02) and for the non-polar residues at protein core (core: 0.05 ± 0.02; surface: 0.04 ± 0.02; interface: 0.05 ± 0.03). These results, albeit not remarked different, support that the core and the interface are the most conserved regions, granting the necessary structural stability at specific PPIs, as previously observed ( 46 ). Additional results are available in the ‘Conservation’ option in the MENSAdb webserver.…”

“…Our main goal with the integration of these features into a single platform is to assist the development of experimental and computational assays, relevant for a better understanding of dimeric MP interactions and interfaces of this largest but poorly studied type of proteins. In the last years, some studies used evolutionary and physicochemical properties similar to the ones provided in our database to train ML for the prediction of MP complex binding sites ( 38 , 46 , 54 ). Nevertheless, as far as we know, herein we offer original features such as the ones from membrane PPI analysis not yet used or provided by other databases more dedicated to MP structures (PDBTM, OPM, MemProtMD and, MPSTRUC) or classification (TCDB).…”

The 3D spatial constraint on 6.1 million amino acid sites in the human proteome

“…To demonstrate the flexibility of our framework to work with custom-built protein 3D structures and to investigate whether COSMIS score could capture constraint imposed by protein-protein oligomerization (Caffrey et al, 2004;Mintseris and Weng, 2005;Li et al, 2019), we compiled a set of 41 potassium ion channel (KCN) genes (Supplementary Table 9) for which variants have been annotated in ClinVar. KCN genes encode proteins that function in homo-oligomeric states (Yu and Catterall, 2004), so we expected interface sites to be under stronger constraint than non-interface sites.…”

Getting to know each other: PPIMem, a novel approach for predicting transmembrane protein-protein complexes

“…Again, most of the approaches are parameterized on soluble globular proteins only, as in [47] , [48] , [49] . Ab-initio prediction of MP interfaces is rendered difficult, as these include amino acid compositions that are not radically different from the rest of the protein surface in terms of average hydrophobicity; nevertheless, they are better conserved than the rest of the surface [50] . This conservation is embedded in the invariable PPIMem nonlinear interface contacting binding motifs that we used.…”