Predicting Disease-Associated N7-Methylguanosine (mG) Sites via Random Walk on Heterogeneous Network.

Recent studies revealed that the modification of N7-methylguanosine (mG) has associations with many human diseases. Effectively identifying disease-associated mG methylation sites would provide crucial clues for disease diagnosis and treatment. Previous studies have developed computational methods to predict disease-associated mG sites based on similarities among mG sites and diseases. However, few have focused on the influence of the known mG-disease association information on calculating similarity measures of mG site and disease, which potentially promotes the identification of the disease-associated mG sites. In this work, we propose а computational method called mGDP-RW to predict mG-disease associations by random walk algorithm. mGDP-RW first incorporates the feature information of mG site and disease with the known mG-disease associations to compute mG site similarity and disease similarity. Then mGDP-RW combines the known mG-disease associations with the computed similarity of mG site and disease to construct a mG-disease heterogeneous network. Finally, mGDP-RW utilizes a two-pass random walk with restart algorithm to find novel mG-disease associations on the heterogeneous network. The experimental results show that our method achieves higher prediction accuracy compared to the existing methods. The study case also demonstrates the effectiveness of mGDP-RW in discovering potential mG-disease associations.