基于LNMF的癌症基因表达谱数据的特征提取  

Feature Extraction for Cancer Gene Expression Profiles Based on Local Non-negative Matrix Factorization

作　　者：王蕊平[1] 王年[1] 苏亮亮[1] 陈乐[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

出　　处：《生物信息学》2011年第2期164-166,170,共4页Chinese Journal of Bioinformatics

基　　金：国家自然科学基金(60772121);安徽大学211工程创新团队项目资助

摘　　要：海量数据的存在是现代信息社会的一大特点,如何在成千上万的基因中有效地选出样本的分类特征对癌症的诊治具有重要意义。采用局部非负矩阵分解方法对癌症基因表达谱数据进行特征提取。首先对基因表达谱数据进行筛选,然后构造局部非负矩阵并对其进行分解得到维数低、能充分表征样本的特征向量,最后用支持向量机对特征向量进行分类。结果表明该方法的可行性和有效性。The existence of mass data is one of the main features of modern information society,and it is very significant for using classification method to analyze tumor gene expression profiles.This paper proposes an algorithm for obtaining feature extraction of tumor gene expression profiles by utilizing local non-negative matrix factorization.The whole process is done by first putting tumor gene expression data into strata.Then,several eigenvectors are obtained by constructing and decomposing local non-negative matrix.These eigenvectors,the dimension of which is low,can fully express the features of the sample data.Finally,the eigenvectors are classified using support vector machine.The feasibility and effectiveness of this algorithm have been proven by experimental results.

关 键 词：分类 特征提取 局部非负矩阵分解 基因表达谱 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]