基于矩阵分解技术的显著基因提取及基因表达数据分析  

Significant Genes Extraction and Analysis of Gene Expression Data Based on Matrix Factorization Techniques

作　　者：孔薇[1] 王娟[1] 牟晓阳 

机构地区：[1]上海海事大学信息工程学院,上海201306 [2]美国罗文大学医药研究中心,新泽西08028

出　　处：《生物医学工程学杂志》2014年第3期662-670,共9页Journal of Biomedical Engineering

基　　金：国家自然科学基金资助项目(61271446);上海市科委青年科技启明星计划(A类)资助项目(11QA1402900);上海市教委科研创新项目资助(11YZ141)

摘　　要：基因之间存在多种多样的表达调控活动,一般认为这些调控关系隐含在基因表达谱中。因此,可以根据基因表达数据对基因调控状态进行建模,以挖掘具有生物学意义的信息及隐含在其中的基因调控关系。本文分别利用独立成分分析(ICA)和非负矩阵分解(NMF)这两种无监督矩阵分解技术对阿尔茨海默病(AD)基因表达数据进行显著基因提取及基因调控网络的构建,通过生物学分析,探讨了两种不同矩阵分解技术在挖掘潜在致病基因上的作用,通过结合两种方法所提取的显著基因的生物学分析,体现了炎症反应在AD致病机制中的重要作用,为AD早期诊断、致病机制研究及基因生物标志物的探寻提供了有益的方法。It is generally considered that various regulatory activities between genes are contained in the gene expres- sion datasets. Therefore, the underlying gene regulatory relationship and the biologically useful information can be found by modeling the gene regulatory network from the gene expression data. In our study, two unsupervised ma- trix faetorization methods, independent component analysis （ICA） and nonnegative matrix factorization （NMF）, were proposed to identify significant genes and model the regulatory network using the microarray gene expression data of Alzheimer＇s disease （AD）. By bio-molecular analyzing of the pathways, the differences between ICA and NMF have been explored and the fact, which the inflammatory reaction is one of the main pathological mechanisms of AD, is also emphasized. It was demonstrated that our study gave a novel and valuable method for the research of ear- ly detection and pathological mechanism, biomarkers＇ findings of AD.

关 键 词：矩阵分解 微阵列基因表达数据 独立成分分析 非负矩阵分解 阿尔茨海默病 

分 类 号：R749.16[医药卫生—神经病学与精神病学]