基于网络拓扑相似性预测潜在致病基因

相关疾病基因的发现和预测是人类基因组研究的重要目标.近些年,一些研究者通过基于网络结构的方法来解决这个难题.然而,大多数方法在推理过程中仅使用了局部的网络信息,并且仅限于推理单一基因的关联.并且这些方法很少考虑到疾病-基因关联网络的网络拓扑性.笔者提出一种改进的基于二部图网络结构推理(improved network-based inference)的计算方法.该方法基于已知的疾病-基因网络拓扑相似性来发现更多潜在致病基因.文中使用的是OMIM数据库中的203种疾病的数据,通过留一交叉验证法验证实验,并获得了88.9%的AUC值.与文中提到的另外两种方法相比,该文方法能够有效地预测潜在致病基因.

Predicting disease genes based on network topological similarity

Discovery and prediction of genes associated with diseases is an important goal of human genome research.In recent years,some researchers have solved this problem through a network-based approach.However,most of these methods that only use local network information in the reasoning process are limited to reasoning single gene association.And these methods rarely take into account the network topology of gene-disease association networks.In this paper,an improved method based on improved bipartite network-based inference is proposed.The method is based on the known gene-disease network topology similarity to discover more potential pathogenic genes.In this study,we used data from 203 diseases in the OMIM database,and validated the experiment by leaving one cross validation method,and obtained an 88.9％ AUC value.Compared with the other two methods mentioned in this article,this method can be effective prediction of potential pathogenic genes.