中文摘要：

提出了一种肿瘤特征基因筛选和基因调控网络的构建方法。首先提取在肿瘤组织和正常组织中表达差别明显的277个基因，通过P-tree决策树方法找出多组肿瘤特征基因，然后结合现有文献中已被实验验证参与了肿瘤形成过程的基因，利用相关系数方法在肿瘤特征基因集中选出与之共通路的基因子集，并用贝叶斯网络方法建立基因闻的调控关系，从而建立对肿瘤特征基因问调控关系的预测模型。

英文摘要：

Research of gene regulatory network is critically important for diagnosis and therapy of tumor. A metnod on construction of gene regulatory network and selection of feature genes for cancer was proposed. Firstly 277 genes were picked up whose expressions differ greatly between tumor samples and normal samples, and multi-group feature genes were collected by method of Peano Count Tree; then the information of literature was combined, the gene was picked which was validated contributing to the process of tumor forming, as the kernel gene of metabolic pathway which it participated in, and the genes were selected sharing the same metabolic pathway from the feature genes by method of correlation coe.ff~cients; then the regulatory relationships of selected genes were estimated by method of Bayesian network, fmally a prediction model of regulatory relationships between tumor genes was constructed.