Firstly, the model obtains four feature representations of drugs from the database, namely, chemical substructure, target, pathway and enzyme, and then calculates the Jaccard similarity of the drugs.