中文摘要：

传统特征基因提取方法往往只注重单个基因在不同样本中的表达差异，忽视了基因之间的关联性以及多个致病基因作为一个基因模块与复杂疾病的联系。针对这种情况，提出基于互信息MI(MutUalInfonmtion)的特征基因提取算法，提取在健康对照和阿尔茨海默症AD(Alzheimer’sdisease)患病样本中关联度具有明显差异变化的基因作为特征基因。在此基础上，结合转录因子TF(Transcriptionfactor)对基因TG(Targetgene)调控的生物学先验信息，利用网络成分分析NCA(NetworkComponentAnalysis)算法分析转录因子的表达活性及其对靶基因的调控强度，构建AD特征基因的转录调控网络。分子生物学分析表明，它们在有丝分裂、细胞周期、免疫反应以及炎症反应过程中的变化对AD的退化过程具有重要作用。

英文摘要：

Traditional feature genes extraction methods tend to focus only on the expression difference of a single gene in different samples,but ignore the correlation among genes as well as the links between multiple pathogenic genes as one gene module and complex diseases. In view of this, we proposed a mutual information-based feature genes extraction algorithm, it is used to extract those genes that have the most significant differences and changes in correlation between the healthy controls and Alzheimer,s disease ( A D ) samples. O n this basis, in combination with the biological priori information about the regulatory of transcription factors ( T F ) on target gene (T G ) , we applied network component analysis algorithm ( N C A ) in analysing T F ’ s expression activities and their regulatory strengths on T G s , and constructed the transcriptional regulatory networks of A D feature genes. Molecular biology analysis showed that the changes of them in mitosis, cell cycle,immune response and inflammation play an important role in deterioration of A D .