Multi-similarity fusion-based label propagation for predicting microbes potentially associated with diseases

More and more scholars have confirmed through research that the microbes in the human body are closely related to human health. Therefore, more and more scholars are committed to researching new prediction methods to discover potential microbe-disease associations. Newly predicted microbes associated with diseases are obtained by a novel way, namely, multi-similarity fusion-based label propagation (MDA-MSFLP). Specifically, the method firstly obtains multiple symmetric similarity matrices about microbes and diseases through calculation. Then, more comprehensive prior information is obtained by the fusion of multiple similarities with the multiple kernel learning (MKL) method. Also, it is considered that the items represented by 0 in the association matrix may be potential associations, so the association matrix is preprocessed according to the fused similarity and the corresponding items represented by 0 will be replaced by the obtained association probability scores. Finally, the label propagation method is used to make predictions for microbes which are potentially related to diseases. To verify its performance, 5-fold cross validation and leave one out cross validation are applied to MDA-MSFLP. It can be found that the results of our method are excellent. Moreover, detailed case studies of three diseases (Chronic Obstructive Pulmonary Disease (COPD), Cystic fibrosis and Psoriasis) are performed. Among the top 15 microbes associated with these three diseases in the predicted results, 10, 11, and 13 have the corresponding literature evidence. It can be concluded that MDA-MSFLP can contribute to the acquisition of microbes associated with diseases.