Figure 2

Re-annotation error rate. Rate of annotation error as a function of the coverage for the re-annotation of the 28088 enzymes of the filtered ENZYME database (See Methods). The full line corresponds to the best-hit strategy (See Determining the optimal correspondence indicator); the curve was obtained by performing the re-annotation for different values of the threshold S₀ between 45 (100% coverage by definition of the filtered ENZYME database) and 841. (∇,∆) correspond to the univariate and multivariate Bayesian methods at the highest confidence level (P = 1, Tables 1 and 2). (◊) corresponds to the "clear cases" identified by the univariate Bayesian method (P = 1 and second highest probability equals to 0; see Re-annotation with the univariate Bayesian approach).