基于Normalized Cut的基因表达数据聚类  

Clustering of gene expression data based on Normalized Cut

作　　者：王俊生[1] 王年[1] 郭秀丽[2] 唐俊[1] 

机构地区：[1]安徽大学计算智能与信号处理教育部重点实验室,安徽合肥230039 [2]山东省信息中心,山东济南250011

出　　处：《安徽大学学报（自然科学版）》2012年第4期68-72,共5页Journal of Anhui University(Natural Science Edition)

基　　金：国家自然科学基金资助项目(60772121);安徽大学"211工程"学术创新团队基金资助项目(KJTD007A)

摘　　要：利用基因表达数据进行聚类分析可提高肿瘤诊断的正确率,对生物医学研究具有重要意义.该文将Normalized Cut应用于基因表达数据的聚类中,将样本映射为高维空间的点,利用亲近矩阵和度矩阵构造正规Laplacian矩阵,经SVD分解得到反映原始样本类别信息的指示向量,利用指示向量各分量的符号差异实现基因表达数据的聚类.通过对白血病和结肠癌数据集的实验,证明了该文方法的有效性.Cluster analysis with gene expression data can improve tumor diagnosis accuracy and it has a positive significance on biomedical research. This paper proposed a method based on Normalized Cut to solve the clustering problem of gene expression data. Firstly, samples were mapped to the points of high dimensional space and normalized laplacian matrix was constructed by affinity matrix and degree matrix. Then, the indicator vector reflecting category information of original normalized laplacian matrix. Finally, the clustering components. The validity of this method was proven samples was obtained by singular value decomposition of problem was solved by using the signs of the indicator vector by experiments on leukemia data and colon cancer data.

关 键 词：聚类 指示向量 Normalized CUT 基因表达数据 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]