Motivation: Most integral membrane proteins form dimeric or oligomeric complexes. Oligomerization is frequently supported by the non-covalent interaction of transmembrane helices. It is currently not clear how many high-affinity transmembrane domains (TMD) exist in a proteome and how specific their interactions are with respect to preferred contacting faces and their underlying residue motifs.
Results: We first identify a threshold of 55% sequence similarity, which demarcates the border between meaningful alignments of TMDs and chance alignments. Clustering the human single-span membrane proteome using this threshold groups ~40% of the TMDs. The homotypic interaction of the TMDs representing the 33 largest clusters was systematically investigated under standardized conditions. The results reveal a broad distribution of relative affinities. High relative affinity frequently coincides with (i) the existence of a preferred helix-helix interface and (ii) sequence specificity as indicated by reduced affinity after mutating conserved residues.
Supplementary information: Supplementary data are available at Bioinformatics online.