决策树算法预测人类病毒的蛋白质磷酸化位点

本文基于决策树分类算法构建人类病毒蛋白质磷酸化修饰位点的预测模型。采用氨基酸物理化学性质对蛋白质序列进行特征提取,并分析丝氨酸、苏氨酸和酪氨酸磷酸化位点邻近序列的氨基酸性质。同时考察了不同分类算法对预测结果的影响。通过10倍交叉验证,利用决策树算法预测丝氨酸、苏氨酸和酪氨酸磷酸化位点的MCC分别达到77.31%、75.91%和71.94%,表明本文提出的方法能有效地预测人类病毒的磷酸化修饰位点。

Predicting Protein Phosphorylation Sites on Human Viruses Based on Decision Tree Algorithm

In this paper, decision tree algorithm is applied to predict the protein phosphorylation sites on human viruses. Physicochemical properties of amino acids are employed to extract the features of protein sequence. In addition, the properties of amino acids around the serine, threonine, tyrosine phosphorylation sites are analyzed and the effects of different classification algorithm on the prediction results are discussed. By decision tree algorithm and the 10-fold cross-validation, the MCCs of serine, threonine and tyrosine phosphorylation sites reach 77.31%, 75.91% and 71.94%., which indicates that the proposed method can effectively predict virus phosphorylation sites.