Prediction of the post-translational modification sites of the protein by machine learning to study the modification mechanism

磷酸化是蛋白质翻译后最重要的修饰之一。利用支持向量机（SVM）的机器学习预测人类蛋白的磷酸化位点。分别构建了功能域和内在无序域的支持向量机来预测目标。结果表明，人类氨基酸序列信息足以对结构域进行有效预测，而特定位点的进化保守性信息则是有效的ID预测。也就是说，在ID中守恒率是不一致的，已知在ID中序列守恒率一般是比较低的。基于同源蛋白新定义了位点特异性保守性，磷酸化位点特别是功能性磷酸化位点的保守率较高，因此对预测是有效的。

Phosphorylation is one of the most important post-translational modifications of the protein. Phosphorylation sites of the human protein are predicted using the machine learning by support vector machine (SVM). SVMs are constructed for prediction target sties in the functional domain and in the intrinsically disordered (ID) region, separately. As the result, human amino acid sequence information is enough for the effective prediction for the domain, while the site specific evolutionary conservation information turns out to be effective for ID. That is, conservation rate is not uniform in ID, where the sequence conservation is known to be relatively low in general. Site specific conservation is newly defined based on the ortholog proteins, and the conservation rate is high at the phosphorylation sites, especially at the functional phosphorylation sites, therefore which is effective for the prediction.