整合蛋白质作用网络结构和序列信息预测药物靶蛋白

Prediction of Drug-Target Proteins by Integrating Protein-Protein Interaction Network and Protein Sequence Similarity

通过整合蛋白质作用网络拓扑结构信息和蛋白质序列信息,对蛋白质作用网络进行聚类,预测药物靶蛋白。分析了网络密度、节点隶属度及蛋白质相对于聚类团的平均相似度,将蛋白质作用网络分割成具有一定生物学特性的子网络,利用维尔克松秩和检验判断聚类团是药物靶点团还是非靶点团。实验结果表明,与仅利用蛋白质作用网络拓扑结构信息的聚类算法相比,该算法预测精度提高17.4%,能够有效预测药物靶蛋白,推测潜在的药物-靶蛋白作用。

In this paper, the authors proposed a novel clustering method to predict drug-target proteins by integrating protein-protein interaction network and protein sequence similarity. By measuring the clustering density, the membership degree and the sequence similarity between a node and a cluster, protein-protein interaction network was divided into clusters with certain molecular biological function. By employing the Wilcoxon rank-sum test, each cluster was judged whether it was a drug-target cluster or a non-target protein cluster. The results showed that the prediction accuracy was 17.4% higher than that of another clustering algorithm without integrating the protein sequence similarity and the proposed method could predict target proteins and infer the potential drug-target interactions.