A novel approach based on bipartite network recommendation and KATZ model to predict potential micro-disease associations

Accumulating evidences indicate that microbes living in the human body plays a very important role in our health and disease, the identification of disease-related microbes will contribute to promote disease biomarker detection and drug discovery for disease treatment, prognosis, diagnosis and prevention. However experimental verification of disease-microbe associations is expensive and time consuming, and there are few methods for predicting potential microbe-disease association. Therefore, developing effective computational models utilizing the wealth of microbe-disease association data collected from previously experimental to identify novel disease-microbe associations is of practical importance. We proposed a novel method based on KATZ model and Bipartite Network Recommendation Algorithm (KATZBNRA) to discover potential associations between microbes and diseases. We compute Gaussian interaction profile kernel of diseases and microbes from known disease-microbe interactions. Then, we construct a bipartite graph and execute bipartite network recommendation algorithm. Finally, we integrate the disease similarity, microbe similarity and bipartite network recommendation score to obtain the final score, which is used to predict novel disease-microbe associations. To evaluate the predictive performance of KATZBNRA, we tested it with the walk length 2 using five-fold cross validation, two-fold cross validation and global leave-one-out cross validation (LOOV), and its AUCs are 0.8969, 0.8463 and 0.9098 respectively. The result showed KATZBNRA is more accurate than two recent similar methods KATZHMDA and BNPMDA.