Predicting the associations between microbes and diseases by integrating multiple data sources and path-based HeteSim scores

Abstract

Abstract A variety of microbial communities are renowned as “a forgotten organ” throughout human body, which have significant impacts on the human health and disease. Identifying the associations between microbes and diseases can provide us with valuable insights for understand the complex disease pathogenesis as well as the diagnosis and therapy, prevention, prognosis drug discovery. Because the experiment-based methods need a long and sampled time series to identify the microbe-disease associations, computational methods provide a valuable insight into understanding complex diseases. However, discovering novel and effective microbial candidates for complex diseases with computational prediction models is still limited. Here, we developed a new method to predict the potential microbe-disease associations by integrating Multiple Data sources and Path-based HeteSim scores for Human Microbe-Disease Associations (MDPH_HMDA). First of all, a heterogeneous network was constructed, in which microbe similarity was measured by the microbe-microbe functional similarity and Gaussian interaction profile kernel similarity for microbes, disease similarity was measured by the symptom-based human disease similarity and Gaussian interaction profile kernel similarity for diseases. Then, normalized HeteSim measure was employed to weight the known microbe-disease pairs, and the HeteSim scores of microbe-disease-disease path and microbe-microbe-disease path were integrated to calculate the relatedness scores for potential microbe-disease associations. Additionally, MDPH_HMDA achieved a reliable prediction performance with AUCs (the area under the ROC curve) of 0.9015 in the leave-one-out cross validation, and the results showed that our method could be effective to find the potential associations between microbes and microbes. Furthermore, representative diseases including type 2 diabetes, colorectal carcinoma, asthma and inflammatory bowel disease (IBD), in which the potential microbes associated with these diseases were ranked as candidate disease-causing microbes, respectively. The reliable performance showed that our proposed method could serve as a powerful computational tool to identify more microbe-disease associations and benefit to medical scientific progress in terms of the human medical improvement.