非线性数据分析方法在基因芯片进行疾病诊断中的应用  

Application of nonlinear data analysis methods in cancer diagnosis with microar-ray data

作　　者：姚微佳[1] 李通化[1] 

机构地区：[1]同济大学化学系,上海200092

出　　处：《计算机与应用化学》2006年第3期214-218,共5页Computers and Applied Chemistry

基　　金：国家自然科学基金资助项目(20275026)

摘　　要：随着人类基因组计划研究的深入,基因芯片技术应运而生,由此为生物医学领域的研究工作带来了新的契机。借助基因芯片技术,成功地对不同类别的癌症正确分类,对于癌症的早期诊断和及时治疗都有重要意义。针对芯片数据少样本多变量的特点,本文创新性地将非线性数据分析方法核映射应用于急性白血病基因芯片数据挖掘工作中,通过构筑核矩阵,在不进行变量选择的前提下,有效地将数据维数大幅度降低,同时利用自组织映射算法,成功地将两类白血病完全分开。此外与使用F-检验进行基因选择的方法进行了比较,最后利用方差分析对数据集的差异性进行了讨论。With the development of microarray technology, biological medicine has stepped into the new era. From the vast volume of array data, vital information can be obtained and extracted, which is critical for cancer classification. The application of this technology is significantly of importance for early diagnosis of illness and timely treatment of it. In this paper, nonlinear data analysis methods, Self-Organizing Map and Kernel Trick, are undertaken to classify two different kinds of leukemia with 100% accuracy. Meanwhile, the results attained from this way are compared with those from normal gene selection method. With the help of Analysis of Variance, the variances of data sets are demonstrated.

关 键 词：基因芯片 白组织映射 核映射 F-检验 方差分析 

分 类 号：Q8-11[生物学]