利用伪氨基酸组分预测分枝杆菌膜蛋白类型

The prediction of types of mycobacterial membrane proteins using pseudo amino acid composition

膜蛋白是重要的药物靶位点,对膜蛋白类型的研究有助于药物的成功设计,因此正确预测膜蛋白类型对于药物研发是十分必要的。本文采用由274条分枝杆菌膜蛋白序列组成的一致性小于40%的数据集,以经过优化的伪氨基酸组分为特征,利用支持向量机分类算法预测分枝杆菌膜蛋白类型,在Jackknife检验下,得到85.4%的总体准确率和72.2%的平均准确率。结果说明,该方法可用于分枝杆菌膜蛋白类型的识别,将有助于抗分枝杆菌药物的开发。

Membrane proteins are targets of many drugs. The study of the types of membrane proteins will help to designing drugs. Therefore, the correct recognition of membrane protein types is essential for drug design. In this study, total of 274 mycobacterial membrane proteins with 〈 40% sequence identity were constructed to train and test our model. By optimizing pseudo amino acid composition, support vector machine was proposed to discriminate the types of membrane proteins of Mycobacterium. Overall accuracy of Jackknife cross - validation achieved 85.4% with the average accuracy of 72.2%. Results show that the method can be used to recognize the type of a newly discovered membrane protein of Mycobacterium, and will be benefit for anti - tubercle bacillus drug design.