Predicting human protein function with multi-task deep neural networks
Table 2
Performance comparison at threshold equal to 0.5 between MTDNN and other prediction methods.
This table reports average F1, Precision, Recall scores of five methods in three GO domains, namely BP, MF, and CC, with the threshold equal to 0.5. MLDNN* is the performance obtained at the thresholds that maximise the F1 scores for individual GO terms in the training set. The best F1 scores in respective domains are marked in bold font.