Quaternary structure predictions of transmembrane proteins starting from the monomer: a docking-based approach

We introduce a computational protocol for effective predictions of the supramolecular organization of integral transmembrane proteins, starting from the monomer. Despite the demonstrated constitutive and functional importance of supramolecular assemblies of transmembrane subunits or proteins, effective tools for structure predictions of such assemblies are still lacking. Our computational approach consists in rigid-body docking samplings, starting from the docking of two identical copies of a given monomer. Each docking run is followed by membrane topology filtering and cluster analysis. Prediction of the native oligomer is therefore accomplished by a number of progressive growing steps, each made of one docking run, filtering and cluster analysis. With this approach, knowledge about the oligomerization status of the protein is required neither for improving sampling nor for the filtering step. Furthermore, there are no size-limitations in the systems under study, which are not limited to the transmembrane domains but include also the water-soluble portions.

Benchmarks of the approach were done on ten homo-oligomeric membrane proteins with known quaternary structure. For all these systems, predictions led to native-like quaternary structures, i.e. with Calpha-RMSDs lower than 2.5 A from the native oligomer, regardless of the resolution of the structural models.

Collectively, the results of this study emphasize the effectiveness of the prediction protocol that will be extensively challenged in quaternary structure predictions of other integral membrane proteins.