Prediction of functional linkages between proteins, based on different methods. (A) Method of domain fusion. The figure shows proteins predicted to interact by the Rosetta stone method (domain fusion). Each protein is shown schematically with boxes representing domains. Proteins P2 and P3 in Genomes 2 and 3 are predicted to interact because their homologues are fused in the first genome. (B) Gene neighbourhood. The figure shows four hypothetical genomes, containing one or more of the genes A, B and C. Since the genes A and B are co-localised in multiple genomes (1–4), they are likely to be functionally linked with one another. (C) Phylogenetic profiles. The figure shows five hypothetical genomes, each containing one or more of the proteins A, B, C and D. The presence or absence of each protein is indicated by 1 or 0, respectively, in the phylogenetic profiles given on the right. Identical profiles are highlighted — proteins A and B are functionally linked (dotted line), whereas proteins C and D, which have different phylogenetic profiles (shown in grey) are not likely to be functionally linked. (D) Correlated mutations. The alignments of two protein families are shown; conserved residues in either alignment are shown in the same colour (blue and green). Correlated mutations in either alignment (coloured red) are indicated by arrow marks. Common sub-trees of the phylogenetic trees are highlighted in yellow. The presence of correlated mutations in each family suggests that the corresponding sites may be involved in mediating interactions between the proteins from each family.