Fig. 2.

Prediction accuracy as estimated by prior knowledge, one round of laboratory validation, and a second iteration of experimental validation. (A) Performance of three function prediction methods and their ensemble as evaluated by a GO-based gold standard. (B) Accuracy of the same predictions as evaluated by a standard augmented with the results of one round of experimental validations (189 tests). (C) New predictions (generated by the same three methods) evaluated using the augmented standard of (B). (D) Accuracy of these predictions evaluated using additional information from a second round of laboratory experiments (52 additional tests). Actual predictive accuracy as evaluated by experimental results is very different than would be expected from a GO-based evaluation.