Weighted SVM在蛋白质磷酸化位点预测中的应用

Weighted SVM for Predicting Phosphorylation Sites

Weighted SVM是标准SVM针对非均衡样本的改进。首次将Weighted SVM用于蛋白质磷酸化位点的预测,在最新版的蛋白质磷酸化数据集PhosphoBase上,取得了目前为止最好的分类精度。k-fold交叉验证和独立测试集实验的结果表明,通过对样本数相对较少的正样本赋予较大的惩罚参数,Weighted SVM有效地改善了分类器向负样本方向的“偏斜”,提高了总的预测正确率以及(正样本)查全率。

Weighted SVM is a modified version of standard SVM for the unhalanced data.This paper first uses Weighted SVM for the prediction of phosphorylation sites,and gets the best performance so far on the lastest version of phosphorylation dataset-PhosphoBase.The k-fold cross validation test and independent dataset test prove that,by using larger penalty parameter to the positive class,which has fewer samples than negative one,in the SVM formulation, Weighted SVM can improve the bias toward negative class effectively,and increase the accuracy and the recall of prediction result.