基于贝叶斯统计方法的两总体基因表达数据分类

在疾病的诊断过程中,对疾病的精确分类是提高诊断准确率和疾病治愈率至 关重要的一个环节,DNA芯片技术的出现使得我们从微观的层次获得与疾病分类及诊断 密切相关的基因功能信息．但是DNA芯片技术得到的基因的表达模式数据具有多变量小 样本特点,使得分类过程极不稳定,因此我们首先筛选出表达模式发生显著性变化的基因 作为特征基因集合以减少变量个数,然后再根据此特征基因集合建立分类器对样本进行分 类．本文运用似然比检验筛选出特征基因,然后基于贝叶斯方法建立了统计分类模型,并 应用马尔科夫链蒙特卡罗(MCMC)抽样方法计算样本归类后验概率．最后我们将此模型 应用到两组真实的DNA芯片数据上,并将样本成功分类．

Classification of Two-class Gene Expression Data Using Bayesian Statistics

During the diagnosis of various kinds of complex diseases, precise classification of diseases into their corresponding category is crucial to improve diagnostic accuracy and probability to heal patients. The advent of DNA-chip technique provides so much information related with gene functions that it makes classification and prediction of diseases under mi-crocosmic level possible. However owing to a small sample size and the large number of variables (genes) in a microarray data set, the classification process is possibly unstable. Therefore this process can be realized by selecting significant differently expressed genes via expression patterns and then classifying samples with these selected features. In this article, likelihood ratio test is applied to select significant differently expressed genes and then a Bayesian classification model is constructed to complete the classification task. In the computation stage we make use of Markov Chain Monte Carlo (MCMC) sampling technique to compute posterior probabilities. In addition the model is applied successfully to two cDNA microarray data sets.