基于KPCA及SVM的蛋白质O-糖基化位点的预测

为了提高蛋白质O-糖基化位点的预测准确率，提出了把核主成分分析（KPCA）与支持向量机（SVM）相结合的方法。实验样本用稀疏编码方式编码，窗口长度为21。首先，用核主成分分析提取了样本的核主成分（特征）；然后，在特征空间中用改进的支持向量机（ISVM）进行分类（预测）。在使用支持向量机分类时，设置了一个边界系数来减少运算的复杂度。实验结果表明，使用KPCA ISVM的方法预测的效果优于PCA SVM的预测效果。预测准确率为87%。更进一步，用不同长度的样本做实验(w=5,7,9,11,21,31,41,51)，使用多数投票法综合各子分类器的优势。结果表明，组合分类器的预测准确率优于子分类器的预测准确率，预测准确率为88%。

Prediction of the protein o-glycosylation by kernel principal component analysis and support vector machines

To improve the prediction accuracy of O-glycosylation sites, a new method of KPCA ISVM was proposed. The samples for experiment were encoded by the sparse coding with window size w=21, kernel principal components(feature) were extracted by kernel principal component analysis(KPCA), then the prediction(classification) was done in feature space by improved support vector machines(ISVM). When using ISVM, a bound coefficient was defined to reduce the complexity of model. The results of experiment show that the performance of KPCA ISVM is better than that of PCA SVM and SVM. The prediction accuracy is about 87%. Furthermore, the same protein sequence under various window size(w=5,7,9,11,21,31,41,51)was investigated, and the majority-vote scheme was used to combine all the pre-classifiers to improve the prediction performance. The results indicate that the performance of ensembles of KPCA ISVM is superior to that of pre-classifier . The prediction accuracy is about 88%.