PT - JOURNAL ARTICLE AU - Thomas A. Hopf AU - Charlotta P.I. Schärfe AU - João P.G.L.M. Rodrigues AU - Anna G. Green AU - Chris Sander AU - Alexandre M.J.J. Bonvin AU - Debora S. Marks TI - Sequence co-evolution gives 3D contacts and structures of protein complexes AID - 10.1101/004762 DP - 2014 Jan 01 TA - bioRxiv PG - 004762 4099 - http://biorxiv.org/content/early/2014/05/06/004762.short 4100 - http://biorxiv.org/content/early/2014/05/06/004762.full AB - High-throughput experiments in bacteria and eukaryotic cells have identified tens of thousands of possible interactions between proteins. This genome-wide view of the protein interaction universe is coarse-grained, whilst fine-grained detail of macro-molecular interactions critically depends on lower throughput, labor-intensive experiments. Computational approaches using measures of residue co-evolution across proteins show promise, but have been limited to specific interactions. Here we present a new generalized method showing that patterns of evolutionary sequence changes across proteins reflect residues that are close in space, and with sufficient accuracy to determine the three-dimensional structure of the protein complexes. We demonstrate that the inferred evolutionary coupling scores distinguish between interacting and non-interacting proteins and the accurate prediction of residue interactions. To illustrate the utility of the method, we predict unknown 3D interactions between subunits of ATP synthase and find results consistent with detailed experimental data. We expect that the method can be generalized to genome-wide interaction predictions at residue resolution.