Fig. 2: Performance metrics in SwissProt E-RXN ASA test set.
From: Multi-modal deep learning enables efficient and accurate annotation of enzymatic active sites
Comparison of a F1 score, b MCC, c Recall, and d FPR between EasIFA and the baseline methods in test set subsets with different levels of sequence identity compared to enzyme sequences in the training set, and comparison of (e) average ROC and (f) average PRC of different EasIFA variants on the test set. The numbers in the square brackets in the figures a–d represent the number of test samples in that subset.
