多发性骨髓瘤基因表达谱分析

为了依据肿瘤基因表达谱数据提取出其中蕴含的样本分类规则，以多发性骨髓瘤的基因表达谱为例，提出了一种在基因表达数据中提取分类特征规则的方法．该方法从统计学角度出发，以基因与样本类别问的相关系数作为衡量属性包含样本分类信息量的标准，并利用神经网络进行属性规约找出分类特征属性集，最后利用决策树进行知识提取，给出样本分类的产生式规则．实验结果表明，所提取出的3条规则对实验样本正确分类率达到100％．

Analysis of Multiply Myeloma Gene Expression Profile

order to extract knowledge for tissue classification from the tumour gene expression profiles, the authors analyzed the gene expression profile of multiply myeloma, and introduced an approach for extracting rules to distinguish different tissue types using statistical method and machine learning approaches. Correlation coefficients of genes with regard to tissue types were used as the criterions for their contribution to classification, and artificial neural networks were employed for feature subset selection. Identified by decision tree algorithm, three classification rules were discovered in the authors experiment, which can distinguish all the tissue types without error.