Figure 2.

BAR 3.0 benchmarking toward the three best performing methods in CAFA2. Besides BAR 3.0, values are as reported in the assessment (8). F1-score is evaluated as the harmonic mean of precision and recall, where precision is the ratio of correct annotation over all the predicted annotation and recall is the ratio of correct annotation over the real annotation. Other methods shown are MS-knn (21), EVEX (22) and the one from Paccanaro Lab. Dashed bars show the upper limits of the performance when exact values are not available. The CAFA2 paper (8) does not list the exact performance of the Paccanaro Lab in CC and BP sub-ontologies and of EVEX on the MF sub-ontology, since the methods did not classify among the best 10 methods. The performance reported with dashed bars correspond to the 10th classified method in the corresponding sub-ontology.