Fig. 3From: DDI-PULearn: a positive-unlabeled learning method for large-scale prediction of drug-drug interactionsThe framework of the proposed method. It consists of the following five components: reliable negative sample identification, feature vector representation for DDIs, PCA compression, DDI prediction, and performance evaluation. RN: reliable negative samples; PCA: principal component analysis; DDI: drug-drug interactionBack to article page