【摘要】 考虑到不同类型的相互作用对于功能预测的作用各不相同,结合蛋白质相互作用网络和蛋白质结构域信息构建多关系蛋白质网络,并为每种类型的相互作用赋予不同的遍历优先级.基于多关系网络,提出一种蛋白质功能预测方法 FPM(Functions prediction based on multi-relational networks).对于未注释的蛋白质,算法遍历与该蛋白质相连的,具有最高优先级的所有相互作用,形成一个候选邻居节点集合.最后根据邻居节点集合形成预测的功能集合,并为每一项功能评分、排序.与其他算法对比结果表明,FPM方法的性能优于其他的功能预测方法.更多还原

【Abstract】 Considering the different influences of different types of interactions in protein function prediction, we construct a multi-relation interaction network by integrating protein-protein interaction(PPI) network and proteins domain information and provide different priorities for different types of interactions. We propose a protein function prediction method, named functions prediction based on multi-relation networks(FPM) to infer protein functions. Given an unannotated protein, all interactions with highest priority associated with the protein are visited to form a candidate neighbor nodes set. Finally, we sort all functions of the neighbor nodes and select part of them to predict an unknown protein.Comparson with other methods show that our method outperforms other function prediction methods.更多还原