Liu Hui, Su Jianzhong, Li Junhua, Liu Hongbo, Lv Jie, Li Boyan, Qiao Hong, Zhang Yan
College of Bioinformatics Science and Technology, Harbin Medical University, Harbin, China.
BMC Syst Biol. 2011 Oct 11;5:158. doi: 10.1186/1752-0509-5-158.
As an important epigenetic modification, DNA methylation plays a crucial role in the development of mammals and in the occurrence of complex diseases. Genes that interact directly or indirectly may have the same or similar functions in the biological processes in which they are involved and together contribute to the related disease phenotypes. The complicated relations between genes can be clearly represented using network theory. A protein-protein interaction (PPI) network offers a platform from which to systematically identify disease-related genes from the relations between genes with similar functions.
We constructed a weighted human PPI network (WHPN) using DNA methylation correlations based on human protein-protein interactions. WHPN represents the relationships of DNA methylation levels in gene pairs for four cancer types. A cancer-associated subnetwork (CASN) was obtained from WHPN by selecting genes associated with seed genes which were known to be methylated in the four cancers. We found that CASN had a more densely connected network community than WHPN, indicating that the genes in CASN were much closer to seed genes. We prioritized 154 potential cancer-related genes with aberrant methylation in CASN by neighborhood-weighting decision rule. A function enrichment analysis for GO and KEGG indicated that the optimized genes were mainly involved in the biological processes of regulating cell apoptosis and programmed cell death. An analysis of expression profiling data revealed that many of the optimized genes were expressed differentially in the four cancers. By examining the PubMed co-citations, we found 43 optimized genes were related with cancers and aberrant methylation, and 10 genes were validated to be methylated aberrantly in cancers. Of 154 optimized genes, 27 were as diagnostic markers and 20 as prognostic markers previously identified in literature for cancers and other complex diseases by searching PubMed manually. We found that 31 of the optimized genes were targeted as drug response markers in DrugBank.
Here we have shown that network theory combined with epigenetic characteristics provides a favorable platform from which to identify cancer-related genes. We prioritized 154 potential cancer-related genes with aberrant methylation that might contribute to the further understanding of cancers.
作为一种重要的表观遗传修饰，DNA甲基化在哺乳动物发育及复杂疾病发生过程中发挥着关键作用。直接或间接相互作用的基因在其所参与的生物学过程中可能具有相同或相似的功能，并共同促成相关疾病表型。基因间的复杂关系可用网络理论清晰呈现。蛋白质-蛋白质相互作用（PPI）网络提供了一个平台，可据此从功能相似基因间的关系中系统识别疾病相关基因。
我们基于人类蛋白质-蛋白质相互作用构建了一个利用DNA甲基化相关性的加权人类PPI网络（WHPN）。WHPN代表了四种癌症类型中基因对的DNA甲基化水平关系。通过选择与已知在这四种癌症中发生甲基化的种子基因相关的基因，从WHPN中获得了一个癌症相关子网（CASN）。我们发现CASN的网络社区连接比WHPN更紧密，这表明CASN中的基因与种子基因的距离更近。我们通过邻域加权决策规则对CASN中154个甲基化异常的潜在癌症相关基因进行了优先级排序。GO和KEGG的功能富集分析表明，优化后的基因主要参与调节细胞凋亡和程序性细胞死亡的生物学过程。对表达谱数据的分析显示，许多优化后的基因在这四种癌症中表达存在差异。通过查阅PubMed共引文献，我们发现43个优化后的基因与癌症及甲基化异常相关，且有10个基因在癌症中被验证存在甲基化异常。在154个优化后的基因中，通过手动检索PubMed，发现有27个曾作为癌症及其他复杂疾病的诊断标志物，20个作为预后标志物在文献中被报道。我们发现31个优化后的基因在DrugBank中被作为药物反应标志物。
我们在此表明，网络理论与表观遗传特征相结合为识别癌症相关基因提供了一个良好的平台。我们对154个甲基化异常的潜在癌症相关基因进行了优先级排序，这可能有助于对癌症的进一步理解。
