Table 3. Sites within a coevolving group are colocated on protein structures.
| gene | group | #sites | Observed, in-group contact density | Expected, in-group contact density | p-value |
|---|---|---|---|---|---|
| atp6 | |||||
| 1 | 64 | 0.37 | 0.33 | 0.0651 | |
| 2 | 66 | 0.5 | 0.34 | <1e-4* | |
| 3 | 46 | 0.54 | 0.23 | <1e-4* | |
| 4 | 17 | 0.09 | 0.08 | 0.3404 | |
| total | 193 | 0.43 | 0.29 | <1e-4† | |
| cytb | |||||
| 1 | 84 | 0.32 | 0.23 | <1e-4* | |
| 2 | 78 | 0.27 | 0.21 | 0.0059* | |
| 3 | 64 | 0.5 | 0.17 | <1e-4* | |
| 4 | 49 | 0.2 | 0.13 | 0.0042* | |
| 5 | 39 | 0.31 | 0.1 | <1e-4* | |
| 6 | 23 | 0.17 | 0.06 | 0.0005* | |
| 7 | 15 | 0.26 | 0.04 | <1e-4* | |
| 8 | 10 | 0.43 | 0.03 | <1e-4* | |
| total | 362 | 0.31 | 0.17 | <1e-4† | |
| cox1 | |||||
| 1 | 130 | 0.28 | 0.16 | <1e-4* | |
| 2 | 116 | 0.25 | 0.14 | <1e-4* | |
| 3 | 103 | 0.49 | 0.12 | <1e-4* | |
| 4 | 52 | 0.26 | 0.06 | <1e-4* | |
| 5 | 49 | 0.24 | 0.05 | <1e-4* | |
| 6 | 17 | 0.03 | 0.02 | 0.1085 | |
| total | 467 | 0.46 | 0.21 | <1e-4† | |
| cox2 | |||||
| 1 | 50 | 0.21 | 0.15 | 0.0037* | |
| 2 | 32 | 0.18 | 0.09 | 0.0005* | |
| 3 | 44 | 0.18 | 0.13 | 0.0056* | |
| 4 | 34 | 0.27 | 0.09 | <1e-4* | |
| 5 | 24 | 0.14 | 0.06 | 0.0021* | |
| 6 | 10 | 0.09 | 0.02 | 0.0039* | |
| total | 194 | 0.32 | 0.19 | <1e-4† | |
| cox3 | |||||
| 1 | 74 | 0.25 | 0.19 | 0.0016* | |
| 2 | 64 | 0.2 | 0.16 | 0.0077* | |
| 3 | 46 | 0.25 | 0.11 | <1e-4* | |
| 4 | 26 | 0.13 | 0.06 | 0.0002* | |
| 5 | 22 | 0.07 | 0.05 | 0.1005 | |
| total | 232 | 0.34 | 0.24 | <1e-4† | |
For each protein, we partitioned the vertices in the coevolution graph into coevolving groups of sites. Contact graph of a protein represents the physical contacts between sites on protein structures. For each protein, for each coevolving group as well as for the entire graph the mean fraction of edges connecting a site with other sites in the same group, the corresponding expected contact density on random partitions of sites, and the corresponding p-values are shown. We applied the Benjamini–Hochberg correction at the 5% alpha level for tests for individual groups. The p-value threshold corresponding to this correction was 0.043.
*, significant under the Benjamini–Hochberg correction for multiple testing (for groups)
†, significant under the Bonferroni correction for multiple testing (for genes).
Next, we asked whether coevolving groups of sites correspond to clusters in the 3D structure of the protein. For this, for each protein, we constructed a second graph, referred to as a contact graph. In this graph, vertices again correspond to sites, but there is just one type of edge: two sites are connected if the minimal distance between heavy atoms of their correspondent residues is under 4Å. Considering each group of sites in the coevolution graph of each protein, we then asked whether the corresponding subgraph is tightly connected in the contact graph.