多特征融合提取算法结合支持向量机预测膜蛋白类型

Prediction of Membrane Protein Types by Combining Features and Weighted-Support Vector Machine

针对膜蛋白分类预测问题,在氨基酸组分基础上引入氨基酸位置信息,计算多种氨基酸残基指数的相关系数并选择最优组合方式;融合2类特征信息对膜蛋白序列进行特征提取;采用支持向量机算法作为分类器,构建了一种新型膜蛋白分类模型,在自检验、Jackknife检验和独立测试集检验3种典型方式下,预测准确率分别为98.25%、88.10%和95.62%.结果表明,多特征融合能够有效提取膜蛋白序列的特征信息,与现有方法相比,该分类模型具有较高的分类预测成功率.

Membrane protein plays a crucial role in cells and makes the material basis for cells to implement various functions. In order to predict the type of membrane protein, which is a crucial fundamental research in the field of the structure and function of membrane protein, this paper introduced the position of amino acid and calculated multi-amino acid index correlation coefficients. Then it constructed a new type of membrane protein classification model that combines two feature classes and support vector machine （SVM）. Under three typical methods （self-consistency, Jackknife and independent dataset）, the accuracy rate of the prediction is 98.25%, 88. 10% and 95.62% respectively. The experimental results demonstrate the usefulness of the above method to extract the characteristic information and predict the type of membrane protein.