基于邻接谱主分量分析的肿瘤分类方法  

Tumor classification based on principal components analysis of adjacency spectrum

作　　者：陈乐[1] 王年[1] 苏亮亮[1] 王蕊平[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：《安徽大学学报（自然科学版）》2011年第4期86-91,共6页Journal of Anhui University(Natural Science Edition)

基　　金：国家自然科学基金资助项目(60772121)

摘　　要：基于谱图理论展开针对基因表达谱数据的分类研究,将反映图结构的特征表示引入到基因表达谱数据分类中,从而高维空间离散点分布问题便可以转化成为具有结构信息的图问题.文中对基因表达谱数据样本点构造高斯权邻接矩阵,SVD分解后,采用特征记分准则进行筛选,找出最大限度区分肿瘤样本与正常样本的主分量作为样本特征,输入KNN分类器进行分类,通过对白血病两个亚型(ALL与AML)与结肠癌表达谱数据进行实验,证明了文中方法的可行性与有效性.Based on spectral graph theory,a gene expression data classification research was proposed.The features that could reflect the graph structures information were introduced into gene expression data classification,so that the discrete tumor sample dots with high dimension could be transformed into the graphs with rich structural information.In this paper,the authors constructed the Gauss adjacency matrix on gene expression samples,and by means of singular value decomposition,found out the main component that could to the utmost distinguish tumor samples and normal samples using feature scoring criteria.Then they put this component as sample characteristics into the KNN classifier.The experiment on leukemia gene expression data（AML and ALL） and colon cancer gene expression data showed the feasibility and effectiveness of this method.

关 键 词：肿瘤分类 主分量分析 邻接矩阵 特征记分准则 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]