College of Computer Science and Electronics Engineering, Hunan University, Changsha, Hunan, 410082, China.
Comput Biol Chem. 2018 Jun;74:142-148. doi: 10.1016/j.compbiolchem.2018.03.022. Epub 2018 Mar 21.
Gene networks are beneficial to identify functional genes that are highly relevant to clinical outcomes. Most of the current methods require information about the interaction of genes or proteins to construct genetic network connection. However, the conclusion of these methods may be bias because of the current incompleteness of human interactome. In this paper, we propose an efficient strategy to use gene expression data and gene mutation data for identifying cancer-related key genes based on graph entropy (iKGGE). Firstly, we construct a gene network using only gene expression data based on the sparse inverse covariance matrix, then, cluster genes use the algorithm of parallel maximal cliques for quickly obtaining a series of subgraphs, and at last, we introduce a novel metric that combine graph entropy and the influence of upstream gene mutations information to measure the impact factors of genes. Testing of the three available cancer datasets shows that our strategy can effectively extract key genes that may play distinct roles in tumorigenesis, and the cancer patient risk groups are well predicted based on key genes.
基因网络有助于识别与临床结果高度相关的功能基因。目前大多数方法都需要基因或蛋白质相互作用的信息来构建遗传网络连接。然而，由于目前人类相互作用组的不完整性，这些方法的结论可能存在偏差。在本文中，我们提出了一种有效的策略，使用基因表达数据和基因突变数据，基于图熵（iKGGE）识别癌症相关的关键基因。首先，我们仅使用基因表达数据基于稀疏逆协方差矩阵构建基因网络，然后，使用并行最大团算法对基因进行聚类，以快速获得一系列子图，最后，我们引入了一种新的度量标准，结合图熵和上游基因突变信息的影响来衡量基因的影响因素。对三个可用的癌症数据集的测试表明，我们的策略可以有效地提取可能在肿瘤发生中发挥不同作用的关键基因，并且可以根据关键基因很好地预测癌症患者的风险组。
