MOESM2 of DDI-PULearn: a positive-unlabeled learning method for large-scale prediction of drug-drug interactions

Additional file 2 This file contains lists of researched drugs, verified DDIs, reliable negative samples generated by DDI-PULearn, and the detailed feature importance ranking results.• Table S1: DDI prediction results using different combinations of drug features.• Table S2: 548 drugs researched in this work.• Table S3: 45,026 reliable negative samples generated by DDI-PULearn.• Table S4: 48,584 verified DDIs in the benchmark dataset.• Table S5: Detailed feature importance ranking results by Random Forrest.• Table S6: 6602 reliable negative sample seeds generated by OCSVM and KNN (XLSX 1,661 kb).