Results of the function prediction tools over a selected test set of 28 proteins.

The reported data refer to annotation precision at the GO level. The starting subset is composed of 62 annotations and predictions have been considered as i) true positive (TP) if the same as the original annotations, ii) more specific (MS) if a more precise functional description was obtained, iii) related terms (RT) if the prediction is either similar to or more generic than the original ones and iv) false positive (FP) if completely wrong.