Figure 3

Rates of orthologue loss (ROL) for E. coli genes annotated as essential are lower than for those annotated as nonessential. ROL is a rate parameter that reflects both the numbers of gene losses and gene gains (see Materials and Methods). A) Histograms of ROL values for nonessential genes in E. coli. B) Histograms of ROL values for essential genes in E. coli. On average, essential genes have a drastically lower ROL than nonessential genes. C) Quantification of the relationship between ROL and essentiality. The receiver operator characteristic (ROC) curve describes the relationship between the fraction of false positives and the fraction of true positives when using ROL as the discriminating variable. A measure of how well ROL predicts essentiality is the area under this curve (the AUC). In this case, the AUC is 0.946; if ROL were perfectly predictive of essentiality, the AUC would be 1.0.