Identification of genes related to proliferative diabetic retinopathy through RWR algorithm based on protein-protein interaction network
Identification of genes related to proliferative diabetic retinopathy through RWR algorithm based on protein-protein interaction network
复制标题
基于蛋白质-蛋白质相互作用网络的RWR算法识别增殖性糖尿病视网膜病变相关基因
DOI：
10.1016/j.bbadis.2017.11.017
复制
发表时间：
2018-06-01
影响因子：
6.2
通讯作者：
Xu, Xun
中科院分区：
文献类型：
--
作者：
Zhang, Jian;Suo, Yan;Xu, Xun
增殖性糖尿病视网膜病变（PDR）是糖尿病最常见的并发症之一，可导致失明。蛋白质组学研究为PDR的发病机制提供了见解，一系列与PDR相关的基因已被识别，但由于实验方法昂贵且耗时，这些基因远未得到充分表征。在我们之前的研究中，我们通过最短路径算法成功识别出35个候选PDR相关基因。在当前研究中，我们开发了一种计算方法，利用带重启的随机游走（RWR）算法和蛋白质 - 蛋白质相互作用（PPI）网络来识别潜在的PDR相关基因。在通过RWR算法获得一些可能的基因后，应用了一个三阶段过滤策略，包括置换检验、相互作用检验和富集检验，以排除由PPI网络结构、较弱的相互作用强度以及基因本体论（GO）术语和生物学通路相似性有限所导致的潜在假阳性。结果，该方法发现了36个候选基因，与我们之前研究中报道的35个基因不同。文献综述表明，这36个基因中有21个得到了先前实验的支持。这些发现表明我们使用不同计算方法所做努力的稳健性和互补效应，从而为研究PDR发病机制提供了一种替代方法。
Proliferative diabetic retinopathy (PDR) is one of the most common complications of diabetes and can lead to blindness. Proteomic studies have provided insight into the pathogenesis of PDR and a series of PDR-related genes has been identified but are far from fully characterized because the experimental methods are expensive and time consuming. In our previous study, we successfully identified 35 candidate PDR-related genes through the shortest-path algorithm. In the current study, we developed a computational method using the random walk with restart (RWR) algorithm and the protein protein interaction (PPI) network to identify potential PDR-related genes. After some possible genes were obtained by the RWR algorithm, a three-stage filtration strategy, which includes the permutation test, interaction test and enrichment test, was applied to exclude potential false positives caused by the structure of PPI network, the poor interaction strength, and the limited similarity on gene ontology (GO) terms and biological pathways. As a result, 36 candidate genes were discovered by the method which was different from the 35 genes reported in our previous study. A literature review showed that 21 of these 36 genes are supported by previous experiments. These findings suggest the robustness and complementary effects of both our efforts using different computational methods, thus providing an alternative method to study PDR pathogenesis.