Predicting transmembrane protein-protein complexes: TransINT, a novel computational approach

Because of their number and diversity, membrane proteins and their mutual complexes represent potential pharmacological targets par excellence for a variety of diseases. This has enormous implications for the design and discovery of new compounds that modulate the interactions between proteins. However, experimental structural data are very scarce and sequence data grows exponentially. To overcome this problem, we devised TransINT, a computational approach for the prediction of higher-order structures of α-helical transmembrane proteins. The approach involved data mining, filtering, and identification and characterization of the amino acid residues at the interface of the complexes as interaction motifs. These latter were expressed in a mathematical formalism used for motif search in databases, and prediction of membrane protein-membrane protein complexes. The number of predicted plasma membrane protein heteromers for the 39 species dealt with is 21,544 for 1,504 membrane proteins. Our template molecular interface-based approach using experimental recognition sites to formulate interaction motifs predicted thousands of binary complexes between plasma membrane proteins across species. The TransINT online database with the annotated predicted interactions is implemented as a web server and can be accessed at https://transint.shinyapps.io/transint/ or http://sertitan.imeve.univ-evry.fr:3838/transint/.