Fig. 5.

Assessment of the quality of predictions of the state-of-the-art software. (A) Percentage of terms from all three domains of GO presented to experimentalists they labeled by a particular descriptor. The ‘unlabeled’ category shows the terms that have not been assigned any designation by the participants; e.g. skipped. Roughly half of each category corresponds to well-researched proteins hemoglobin and p53, whereas others correspond to the two specific proteins that each individual studies experimentally. (B) Percentage of experimentalists characterizing the prediction of state-of-the-art protein function predictors in response to the survey question ‘Do you think the algorithm has done a good job?’