Inference of direct residue contacts in two-component signaling

Since the onset of the genomic era more than 1000 bacterial genomes have been sequenced and several fold more are expected to be completed in the near future. These genome sequences supply a wealth of information that can be exploited by statistical methods to gain significant insights into cellular processes. In Volume 422 of Methods in Enzymology we described a covariance-based method, which was able to identify coevolving residue pairs between the ubiquitous bacterial two-component signal transduction proteins, the sensor kinase and the response regulator. Such residue position pairs supply interaction specificity in the light of highly amplified but structurally conserved two-component systems in a typical bacterium and are enriched with interaction surface residue pairings. In this chapter we describe an extended version of this method, termed "direct coupling analysis" (DCA), which greatly enhances the predictive power of traditional covariance analysis. DCA introduces a statistical inference step to covariance analysis, which allows to distinguish coevolution patterns introduced by direct correlations between two-residue positions, from those patterns that arise via indirect correlations, that is, correlations that are introduced by covariance with other residues in the respective proteins. This method was shown to reliably identify residue positions in spatial proximity within a protein or at the interface between two interaction partners. It is the goal of this chapter to allow an experienced programmer to reproduce our techniques and results so that DCA can soon be applied to new targets.