Predicting Microbe-Disease Association by Kernelized Bayesian Matrix Factorization

Abstract

The study of microbe-disease associations can be utilized as a valuable material for understanding disease pathogenesis. Developing a highly accurate algorithm model for predicting disease-related microbes will provide a basis for targeted treatment of the disease. In this paper, we propose an approach based on Kernelized Bayesian Matrix Factorization (KBMF) to predict microbe-disease association, based on the Gaussian interaction profile kernel similarity for microbes and diseases. The prediction performance of the method was evaluated by five-fold cross validation. KBMF achieved reliable results which is better than several state-of-the-art methods with around 8% improvement of AUC. Furthermore, case studies have demonstrated the reliability of the method.