Figure 1

Distribution of ROC areas under curves for E. coli essential gene prediction using CMIM selected subsets. Probability density functions of the areas under curves from ROC analysis of 100,000 randomly selected subsets of five organisms selected from a total of 180 sequenced prokaryotes (blue) and all subsets of 5 organisms selected from a set of 20 which were in turn selected based on the conditional mutual information selection procedure described in the text (red). Sensitivity and specificity were computed based on the numbers of true positives (essential genes) and false positives (non-essential genes) identified as the number of organisms in which an ortholog of a given E. coli gene is present in decreases.