Drug-Drug Network Based on Similarity of Molecular Fingerprint

We built a network of FDAapproved small drugs, according to the similarity of molecular fingerprint. With the therapeutic function indicated in drug ATC classification system, we propose a method to predict the new function of drugs based on the drugdrug network. 1 277 Drugtherapeutic function pairs were predicted, about 59.44% of them were supported by scientific literatures. Predicted functions of 55.89% drugs with known ATC code were wellmatched with their approved therapeutic function. While, the unmathed drugtherapeutic pairs may indicate new functions of drugs.