【作者】 李建更； 段艳华； 阮晓钢；

【Author】 LI Jian-geng,DUAN Yan-hua,RUAN Xiao-gang(School of Electronic Information and Control Engineering,Beijing University of Technology,Beijing 100022,China)

【机构】 北京工业大学电子信息与控制工程学院；

【摘要】 结合基因过滤和启发式k-均值聚类分析,提出一种提取肿瘤分类特征基因的新方法。首先对“信噪比”公式进行改进,对基因排序过滤得到信息基因;然后对信息基因进行启发式k-均值聚类分析降低冗余度,选取每类中到所有基因距离总和最小的基因作为这类的特征基因;汇合每类的特征基因作为样本的分类特征。采用所提出的方法对急性白血病基因表达谱进行分析,结果表明所提出的方法比别方法的特征基因冗余度小,且分类效果相当或更好。更多还原

【Abstract】 Combining gene ranking with heuristic k-means clustering analysis,a new hybrid method was proposed.Firstly,the improved “signal to noise” approach was applied to select a set of top-ranked genes;secondly,a heuristic k-means clustering algorithm was applied to analyze these genes for reducing the redundancy;finally,a representative gene,which is the gene with minimum sum of squares of distance to all other genes from each cluster,was selected;then the set of representative genes was considered classification features.This method was used to analyze the acute leukemia gene expression profiles data.The experimental result reveals that the proposed approach can achieve a smaller redundancy with the same or better classification accuracy.更多还原