摘要
在构建蛋白质相互作用网络时从通路数据出发,利用通路数据整合蛋白质相互作用网络。提出了一个基于随机游走的蛋白质功能预测方法,在该算法中把已知功能的蛋白质当作起始点,对于随机游走在蛋白质互作网络中产生的邻居信息,转换为注释模式信息,并根据已知功能的蛋白质的功能来提取未知功能蛋白质的注释模式。利用传统的K近邻算法从训练样本集中找到未知功能蛋白质的k个最近邻。最后,结合多标签分类的K近邻算法,统计k个近邻中蛋白质的功能类数目,基于最大后验概率预测未知功能蛋白质属于的功能标签类。通过在构建蛋白质互作网络进行实验,结果表明提出的方案能够有效地进行蛋白质功能预测。
In the process of constructing the protein-protein interaction network is starting from the pathway data, data integration using pathway of protein interaction networks. A prediction method for protein function is presented based on random walk, in this algorithm, the proteins of known function as a starting point is provided, for the random walk in the protein interaction network of neighborhood information, annotation schema information is converted, and according to the known function protein function to extract the unknown function protein annotation mode. After, using the traditional KNN algorithm to find the unknown function protein the k nearest neighbors from the training set. Last, combined with KNN algorithm for multi label classification, the maximum posteriori probability prediction function label unknown function protein is based. The experiments in which we construct protein interaction network, the results show that the proposed scheme can predict the protein function effectively.
出处
《黑龙江大学工程学报》
2015年第3期73-78,共6页
Journal of Engineering of Heilongjiang University
基金
黑龙江省自然科学基金资助项目(F201248)
