A Novel Approach Based on Bi-Random Walk to Predict Microbe-Disease Associations

Abstract

An increasing number of clinical observations have confirmed that the microbes inhabiting in human body have critical impacts on the progression of human disease, which provides promising insights into understanding the mechanism of diseases. However, the known microbe-disease associations remain limited. So, we proposed Bi-Random Walk based on Multiple Path (BiRWMP) to predict microbe-disease associations. Leave-one-out cross-validation (LOOCV) and 5-fold cross-validation were adopted to demonstrate the capability of proposed method. BiRWMP performed better than other methods. Finally, we listed 2 common disease and potential microbes ranked at top 10, and we demonstrated its reasonableness through looking up literatures.