Identifying New Methylated Arginines via Granular Decision Fusion with SVM Modeling

Traditional methods of discovering new methylated arginines in proteins involve conducting delicate experiments to examine every arginine in the primary sequence. Such process is labor-intensive and time-consuming. To speed up this process, one popular way is using machine learning method to model the underlying mechanism of protein methylation based on known methylated proteins and then suggesting candidate positions on unknown proteins. In this paper, we first collect several proteins methylated by different families of PRMTs, and then use granular computing methods to build a granular decision fusion method based on SVM modeling. Such decision fusion method can produce high prediction accuracy. More importantly, we use this method to successfully discover several highly possible methylation sites on some unknown proteins, biological experiments have verified our results.