Results from phylogenetic profile comparison of 635,628 pairs of proteins chosen from among a subset of 1,191 yeast proteins. Plot representations are same as in Figure 3. (a) Predictive power of pyholgenetic profile analysis. The relative performances of all but one (BAE4) reference sets of genomes for yeast are almost the same as that for E. coli. A gradual decrease in performance can be noticed as the fraction of eukaryotic genomes in the reference set increases (BAE4, BAE5, BAE6, AE, E), which is counter-intuitive considering the fact that one would hope to see an improved performance on adding more eukaryotic genomes. This suggests that the diverse physiology of eukaryotes could be adding substantial noise to the protein profiles making it difficult for the method to separate functionally related protein pairs from unrelated protein pairs. Use of bacterial and archaeal genomes (BA) alone results in a good performance, a result that lends support to serial endosymbiosis theory. (b) Sensitivity versus specificity plot.