Gene Selection with Fold Change Detection via Nonnegative Matriz Factorization
Fold change is a commonly used metric to measure differentially expressed genes in microarray data analysis. There are lots of methods for calculating the fold change from statistical methods to many complex and computation demanding approaches. However, most of them need to know the sample labels, especially in the case of multiple samples. This paper provides a new method to detect the fold change without sample labels. Based on the assumption that we only know two types of distinct conditions, the method exploits nonnegative matrix factorization 1 for reducing gene expression data into two metagenes, each of which stands for a representative sample for a family of samples under same condition 2. The estimated fold change between metagenes can be used for gene ranking or selection. The experimental results on two real cancer data sets show that most of top 10 genes we discovered are consistent with the results via RankGene 3; furthermore, some new informed genes have biological significance, and they should be considered in further biological investigation with superiority. The proposed method can work well without sample labels.
Weixiang Liu Tianfu Wang Siping Chen Aifa Tang Kehong Yuan Datian Ye
Shenzhen Key lab of Biomedical Engineering School of Information Engineering Shenzhen University, Sh Shenzhen Key Lab of Male Reproduction and Genetics Peking University Shenzhen Hospital Shenzhen, 518 Life Science Division, Graduate School at Shenzhen Tsinghua University, Shenzhen, 5180S5, China
国际会议
The 7th Asia-Pacific Bioinformatics Conference(第七届亚太生物信息学大会)
北京
英文
838
2009-01-01（万方平台首次上网日期，不代表论文的发表时间）