基于R软件的Lasso回归在肿瘤信息基因选择中的应用  

The Application of Lasso Regression to Tumor Informative Gene Selection Based on R Software

作　　者：徐庆娟[1] 杨彬彬[1] 

机构地区：[1]广西师范学院数学与统计科学学院,广西南宁530023

出　　处：《广西师范学院学报（自然科学版）》2016年第4期36-41,共6页Journal of Guangxi Teachers Education University(Natural Science Edition)

基　　金：混合与缺失数据统计分析广西高校重点实验室科学基金开放项目(GXMMSL201407);广西高校科学技术研究项目(KY2015YB190)

摘　　要：如何从肿瘤基因表达谱成千上万个基因中提取与疾病相关的信息基因,已成为肿瘤分类问题的研究核心.该文介绍了Lasso回归,以公开的结肠癌数据集为分析对象,采用信噪比指标对基因排序过滤无关基因.然后利用R软件中求解Lasso的程序包msgps和glmnet进行降维,从而选择出信息基因.与前人研究结果比较,说明了Lasso回归的有效性.该文的算法利用R软件实现,代码公开,操作简单.How to extract informative gene associated with tumor from thousands of gene in tumor gene expression profile, has been the core of the research of tumor classification. In this paper, the lasso regression is introduced firstly. The open colon cancer data set is taken as the analysis object, and a large number of genes is filtered by ranking the value of signal noise ratio. Then the packages of msgps and glmnet of R software solving Lasso regression are respectively used to reduce dimensionality of the gene data filtered, and the informative gene is gained. Compared with previous research results, the results of this paper is effective based on Lasso regression and R software. The algorithm of this paper is implemented with R language, in which the code is open and easy to operate.

关 键 词：Lasso 基因表达谱 基因选择 R软件 

分 类 号：O212.4[理学—概率论与数理统计]