Figure 2

Hierarchical evaluation measures. Boxed classes correspond to annotations and filled classes to predictions. n _p : number of proteins with at least one prediction (correct or not), $n_p^{+}$: number of correct predictions, $n_p^{-}$: number of incorrect predictions, $n_p^{\ast }$: number of missing predictions, n is the number of proteins to be annotated, pr is the fraction of proteins with an annotation.