Prediction of Protein Sumoylation Site

[[abstract]]Protein sumoylation is a post-translational modification that plays the key role in biological process. According to reports, sumoylation proteins involve in a wide variety of cellular activities and closely associated with many diseases. However, it is very difficult to identify the protein sumoylation sites by experimental approaches. Using computational methods to find out the protein sumoylation sites will be able to further understand its mechanism and improve the experimental design.In this study, the protein sumoylation sites were predicted by combining the support vector machine and genetic algorithms with feature selection. In our dataset 1, there were 144 proteins which contained 250 sumoylation sites and 5,742 non-sumoylation. 174 sumoylation sites had the consensus motif, ΨKXD/E and were removed from the dataset 1 as the dataset 2. Our method achieved the 97.83% accuracy, 81.58% sensitivity, 98.36% specificity and 0.70 Matthews Correlation Coefficient in the dataset 1. On the other hand, the sumoylation sites which lacked of the consensus motif in the dataset 2 were also predicted well by 99.48% accuracy, 85.93% sensitivity, 99.63% specificity and 0.79 Matthews Correlation Coefficient. It showed a significant improvement of prediction in sumoylation sites with and without the consensus motif