基于ICASSO的乳腺癌基因表达数据可靠特征提取

Reliability feature extraction of breast cancer gene expression data based on ICASSO

运用独立成分分析(ICA)方法对乳腺癌基因表达谱数据进行特征提取,并采用聚类及可视化的可靠性评估方法(ICASSO)对所提取的独立成分(IC)进行评估和综合可得到与疾病更加密切相关的显著基因。结果显示,通过与乳腺癌发病的相关性分析,选取显著性高的IC分析发现,其100个特征基因当中,经分子生物学实验验证与乳腺癌发病密切相关的基因有35个,在此基础上通过从基因调控网络的角度分析这些特征基因所参与的生物过程,进一步证实了部分生物过程也与乳腺癌发病密切相关。

Using correlation analysis and feature vector clustering method （ investigating the reliability of ICA, ICASSO）, independent component analysis method for breast cancer gene expression profile data feature extraction is evaluated for reliability. The significant genes more closely related to disease by evaluation and integration of extracted independent component （IC） could be obtained. The results showed that through correlation analysis of breast cancer, the significant IC were found. There were 35 genes that were confirmed by experiment of molecular biology closely related with breast cancer among the 100 feature genes of the significant IC. On that basis, from the point of view of gene regulatory network analysis, the biological processes that the feature genes were involved were found. And it is further confirmed that part of the biological process is also closely related with breast cancer.