National Engineering Laboratory for Cereal Fermentation Technology, Jiangnan University, Wuxi, China.
The Key Laboratory of Industrial Biotechnology, Ministry of Education, School of Biotechnology, Jiangnan University, Wuxi, China.
Sci Rep. 2017 Jul 28;7(1):6827. doi: 10.1038/s41598-017-07189-6.
Breast cancer encompasses a group of heterogeneous diseases, each associated with distinct clinical implications. Dozens of molecular biomarkers capable of categorizing tumors into clinically relevant subgroups have been proposed which, though considerably contribute in precision medicine, complicate our understandings toward breast cancer subtyping and its clinical translation. To decipher the networking of markers with diagnostic roles on breast carcinomas, we constructed the diagnostic networks by incorporating 6 publically available gene expression datasets with protein interaction data retrieved from BioGRID on previously identified 1015 genes with breast cancer subtyping roles. The Greedy algorithm and mutual information were used to construct the integrated diagnostic network, resulting in 37 genes enclosing 43 interactions. Four genes, FAM134B, KIF2C, ALCAM, KIF1A, were identified having comparable subtyping efficacies with the initial 1015 genes evaluated by hierarchical clustering and cross validations that deploy support vector machine and k nearest neighbor algorithms. Pathway, Gene Ontology, and proliferation marker enrichment analyses collectively suggest 5 primary cancer hallmarks driving breast cancer differentiation, with those contributing to uncontrolled proliferation being the most prominent. Our results propose a 37-gene integrated diagnostic network implicating 5 cancer hallmarks that drives breast cancer heterogeneity and, in particular, a 4-gene panel with clinical diagnostic translation potential.
乳腺癌包括一组异质疾病，每种疾病都与不同的临床意义相关。已经提出了数十种能够将肿瘤分类为具有临床相关性亚组的分子生物标志物，尽管这些标志物在精准医学中做出了重要贡献，但它们也使我们对乳腺癌亚分型及其临床转化的理解变得更加复杂。为了解码具有乳腺癌诊断作用的标志物网络，我们通过整合 6 个公共基因表达数据集和从 BioGRID 检索到的蛋白质相互作用数据，构建了诊断网络，这些数据是基于先前确定的 1015 个具有乳腺癌亚分型作用的基因。我们使用贪婪算法和互信息构建了综合诊断网络，结果包含 43 个相互作用的 37 个基因。通过层次聚类和支持向量机和 k 最近邻算法的交叉验证，我们发现 4 个基因（FAM134B、KIF2C、ALCAM、KIF1A）与最初的 1015 个基因具有可比的亚分型效果，具有相似的亚分型效果。通路、基因本体和增殖标志物富集分析共同表明，有 5 个主要的癌症特征标志驱动着乳腺癌的分化，其中促进不受控制的增殖是最显著的。我们的研究结果提出了一个由 37 个基因组成的综合诊断网络，其中包含 5 个癌症特征标志，这些标志驱动着乳腺癌的异质性，特别是一个具有临床诊断转化潜力的 4 基因面板。
