基于提升小波的基因芯片数据的分类预测  

Gene microarray data classification and prediction based on lifting wavelet

作　　者：凌玲[1] 衣娜[1] 王翼飞[1] 

机构地区：[1]上海大学理学院,上海200444

出　　处：《应用数学与计算数学学报》2014年第2期218-227,共10页Communication on Applied Mathematics and Computation

基　　金：国家自然科学基金资助项目(30971480)

摘　　要：针对肿瘤的早期诊断,提出了一种基于提升小波变换的特征提取的方法,对肿瘤数据样本进行分析鉴别.该方法利用提升小波变换对190例肝癌(包括对照)和107例肺癌(包括对照)基因表达谱芯片数据进行处理后,提取信号的低频信息,经支持向量机训练学习,构造分类器模型,用于癌和非癌样本的区分甄别.实验结果表明,经提升小波变换提取的特征基因,送入分类器中能得到较高的分类率,且在支持向量机中选取线性核函数或径向基函数都能达到较好的分类效果.通过随机选取的20例基因表改谱芯片样本,对所建立的模型进行了测试,获得了很好的效果,因此,本文提出的方法对肿瘤的诊断有一定的应用意义.For the problem of the early detection of cancer, a method of feature extraction based on the lifting wavelet is introduced to analyze and identify tumor samples. With this method, the 190 liver cancer gene expression profiles samples （including control group） and 107 lung gene expression profiles samples （including control group） are calculated by the lifting wavelet, and the low frequency information is extracted as features. These features then are learned by the support vector machine （SVM） to train a model for distinguishing the cancer and noncancer samples. Numerical results report that, the feature genes which extracted by the lifting wavelet transform can get a high classification rate after sending into the classifier. Results also indicate that both the linear kernel function and the radial basis function （RBF） selected as the kernel function in the SVM can reach an ideal classification effect. The model is tested with 20 gene expression profiles samples which are chosen at random, and it offers perfect performance. Therefore, the method presented in this paper has practical value for the diagnosis of tumors.

关 键 词：提升小波 支持向量机 核函数 交叉验证 肝癌和肺癌基因芯片 

分 类 号：R730.4[医药卫生—肿瘤] TP18[医药卫生—临床医学]