Table 2. Performance for each feature set by 10-fold cross-validation.

Feature set	Sensitivity	Specificity	Accuracy	AUC
All	88.6	88.4	88.4	0.945
88	89.5	88.7	88.7	0.951
70	91.3	88.3	88.3	0.952
34	91.1	88.8	88.8	0.954

Here, the first 34, 70 and 88 features yielding the greatest contributions were selected to construct the prediction model for catalytic residues.