Figure 1.

Confidence intervals generated from GOcardShuffle. Confidence intervals for the correlation of evolutionary rates and expression levels (mRNA expression levels are used as proxies for protein abundance). Incorporating GO annotations, individually, in pairs, or all three categories together results in progressive right-shifts of the distribution under the conditional Null models. The real data is indicated by the red vertical lines. When all three categories of biological information are included the distribution obtained under the Null model covers the observed correlation for both evolutionary rates and expression levels of interacting proteins. Function, Process and Compartment are indicated by F, P and C, respectively (the approximation Eqn. (12) was used to calculate the weight matrices for multiple annotation categories; for the yeast dataset used here this appears to be a reasonable approximation).