基于微阵列基因表达谱的一种关联空间的癌症分类算法

A Relative Space Based Cancer Classification with Gene Expression Profiles

利用微阵列基因表达谱分类癌症患者样本对患者的治疗具有非常重要的意义.针对高维、高冗余的微阵列基因数据中致癌因子存在局部相关性的特点,提出一种基于权重的关联空间分类模型(Weight based Classification with Related Space,WCRS).基本思想是首先利用协方差矩阵的对角化来构建癌症组的关联空间,并提出一种基于癌症关联空间的基因表达模式,然后提取使得癌症组具有最小组能量的最小扩展空间,最后在最小扩展空间上建立一种基于权重的癌症分类算法.实验结果表明,WCRS在精确度上比传统分类算法具有更好的性能.

Classification of patient samples with gene expression profiles is important to cancer treatment. In the large redundant and high dimensional gene expression data, a cancer is sensitive to some cancerogehic factors while another cancer is sensitive to some others. So we proposed a weight based classification with relative space（WCRS）. The main idea is that a cancer＇s relative space is obtained via the diagonalization of its covariance matrix, and we built the cancer＇s model based on its relative space. Then the energy of a cancer is presented for treasuring its relative spaces, and a minimal spread space based classification algorithm is proposed. The experiments show WCRS makes better precision than traditional classifications.