Feature Threshold Recall (%) Precision (%) Accuracy (%) MCC AUC B-factor 0.11 55.6 18.6 71.8 0.195 0.694 RASA 0.07 84.8 11.4 46.9 0.154 0.683 B-factor+RASA 0.14 54.7 28.9 85.0 0.317 0.797

Table 2: The performance of fivefold cross-validation on Andersen et al.’s dataset.