Fig. 1

The framework of this work. Firstly, multi-sources of data, such as drug substructures, disease phenotypes, protein-protein interactions (PPI), gene profiles and network profiles, are well organized into three layers. Secondly, those information are utilized to calculate similarity scores as well as drug-disease treatment priori. Thirdly, the similarity matrices and the priori are integrated to construct drug-disease pair graph. Finally, SSGC algorithm is applied to predict drug-disease treatment relations