Growing evidence shows that microbes in human body and body surface play critical roles in the development of many human diseases. Predicting the underlying associations between diseases and microbes is essential for deeply understanding the pathogenesis of diseases. However, biological experiments to find the relationship between microbes and diseases is usually laborious and time-consuming, which presents the need for effective computational tools. In this study, we propose a computational model of node-information-based Link Propagation for Human Microbe-Disease Association prediction (LPHMDA) to prioritize disease-related microbes. LPHMDA and 3 popular methods including KATZHMDA, PBHMDA, and LRLSHMDA were implemented and compared on the Human Microbe-Disease Association Database (HMDAD) based on cross-validation. As a result, LPHMDA achieved an AUC of 0.9135 in leave-one-out cross-validation (LOOCV), outperforming those of the 3 compared methods. In addition, the performances of LPHDMA on the 3-fold CV, 5-fold CV and 10-fold CV were also better than those of the other 3 canonical methods, further demonstrating its superiority. Finally, we took colorectal carcinoma, asthma and obesity as case studies. Interestingly, 9, 9 and 8 of the top 10 novel microbes predicted by LPHMDA to be associated with the 3 diseases respectively could be confirmed by literatures, providing potential disease-associated microbes for further experimental validation. In summary, LPHMDA is an effective method for prioritizing disease-associated microbes.
- link propagation
- microbe-disease association