基于支持向量机的多类凋亡蛋白亚细胞位置预测

Prediction of Multi-Class Subcellular Locations of Apoptosis Proteins Using Support Vector Machines

基于支持向量机,以全部和局部氨基酸序列的n肽组分、序列的亲疏水性分布等五种特征提取方法构成特征向量表示蛋白质序列,对六类细胞凋亡蛋白的亚细胞位置进行预测.结果表明,基于氨基酸二肽组成成分构成的特征向量集(以符号DIPE表示)的预测结果高于其它四种特征向量集的预测结果,在Jackknife检验下,总预测成功率达到了89.3%;与现有的方法比较,发现对于Mitochondrial类凋亡蛋白,支持向量机方法有更好的预测效果.

The six kinds of subcellular locations of apoptosis proteins are predicted by using support vector machine algorithm based on the n-peptide components of global and partial amino acid sequence as well as the distribution of hydropathy along the amino acid sequences. The predictive results for different parameters show that the overall prediction accuracy is better by using of the dipeptide composition （DIPE）. In the jackknife test, the overall prediction accuracy reaches 89.3%. Compared with the other existing algorithms,the SVM algorithm can give the best prediction quality for the Mitochondrial subcellular locations of apoptosis proteins by the jackknife test.