基于结构域的芥子油苷合成相关蛋白质相互作用模型构建和预测

Domain Based Prediction of Glucosinolates Biosynthetic Pathway Related PPIs

从蛋白质一级序列抽取蛋白质结构域信息建立结构域特征向量,使用结构域特征向量训练基于高斯径向基核的支撑向量机,用5折交叉验证对支撑向量机参数网格寻优,然后用最优化参数训练得到蛋白质相互作用预测器。结果表明:预测器的交叉验证ROC曲线下面积达到了85.49%,可以有效预测拟南芥相关的蛋白质相互作用。

We extracted the protein domain information from the first level structure, and constructed the feature vectors by the domain information. A support vector machine （SVM） was trained by the feature vectors, and the five-fold eross-validation was used to conduct the grid optimization for the SVM parameters to get the protein-protein interaction predictor trained by the optimum parameters. The area under ROC of cross-validation reached 85.49 with good effectiveness of the predictor.