Prediction of protein phosphorylation sites by support vector machines

Protein phosphorylation is one of the most important post-translational modifications, and revealing its mechanism is an important research topic. In this paper, phosphorylation sites in human proteins are predicted by support vector machine (SVM). First, two types of SVMs are constructed, each for phosphorylation sites in domain and in intrinsically disordered region (IDR). In domain, wide range of information of amino acid sequence is found effective, while it is not effective in IDR. As phosphorylation is abundant in IDR, the second part of the study focuses on the prediction of phosphorylation sites in IDR, especially, the phosphorylation sites with any known function. Then, it is found that the evolutionary conservation of each site is different in IDR, and multiple ortholog sequences which contain the conservation information is effective for the prediction compared with single sequence information.