Mining Multi Drug-Pathways via A Probabilistic Heterogeneous Network Multi-label Classifier

Mining drug networks is a very important research issue to discover hidden relations between multi drug-entities relations, such as multi drug-pathways, multi drug-targets, and multi drug-diseases. One very important relation is the drug-pathway, where drugs affect the human body through their pathways. In this paper, a probabilistic Heterogeneous Network Multi-label Classifier (HNMC) is proposed to classify multi drug-pathways relations. Data is collected from Drugbank.ca [1], Kegg (keg drug, Kegg diseases, Kegg pathways, Kegg orthologs, Kegg brite) [2] and small molecular pathways [3,4]. For drug-pathways data, two datasets are considered: one is based on Drug-Drug Interaction (DDI) and the other is based on Drug-Pathways Interactions (DPI). HNMC has proved its efficiency with an average of 90% precision, 92.35% recall, 92% accuracy, and 96% ROC