Background: Drug repositioning refers to discovering new indications for the
existing drugs, which can improve the efficiency of drug research and development.
Methods: In this work, a novel drug repositioning approach based on integrative
multiple similarity measure, called DR_IMSM, is proposed. The process of integrative
similarity measure contains three steps. First, a heterogeneous network can be
constructed based on known drug-disease association, shared entities information for
drug pairwise and diseases pairwise. Second, a deep learning method, DeepWalk, is
used to capture the topology similarity for drug and disease. Third, a similarity
integration and adjusting process is further conducted to obtain more comprehensive
drug and disease similarity measure, respectively.
Results: On this basis, a Bi-random walk algorithm is implemented in the constructed
heterogeneous network to rank diseases for each drug. Compared with other
approaches, the proposed DR_IMSM can achieve superior performance in terms of
AUC on the gold standard datasets. Case studies further confirm the practical
significance of DR_IMSM.