从肿瘤基因表达数据挖掘分类规则的研究  

Mining Classifying Rules from Tumor Gene Expression Data

作　　者：马猛[1,2] 汝颖[3] 马腾[4] 钮俊清[2] 李龙澍[1] 王煦法[2] 

机构地区：[1]安徽大学计算机科学与技术学院,合肥230039 [2]中国科学技术大学计算机科学与技术系,合肥230027 [3]上海交通大学医学院,上海200025 [4]安徽省立医院心血管内科,合肥230001

出　　处：《中国生物医学工程学报》2009年第5期707-712,共6页Chinese Journal of Biomedical Engineering

基　　金：安徽省自然科学基金(050420204);安徽省高校学科拔尖人才基金(05025102);安徽省高校青年教师基金(2006jql038);安徽省科技攻关计划重大科技专项项目(8010201002)

摘　　要：基于肿瘤基因表达数据,利用信息科学的方法和技术建立肿瘤预测分类模型,对肿瘤基因表达模式研究和肿瘤的诊断识别具有重要意义。本研究提出一种从肿瘤基因表达数据中直接挖掘分类规则建立肿瘤预测分类器的方法。该方法首先抽取实验样本集,分别找出标记肿瘤和正常组织样本的分类特征,由此生成可预测样本类别的分类规则,对每个未知类别样本,按照置信度最高原则,选择一个分类规则作为预测结构。本研究的实验数据来自Broad Institute的前列腺癌基因表达数据,实验结果显示该方法的预测精度在90%以上,且同时获得了大量结构透明的分类预测规则,表明本研究的方法是可行的和有效的。Establishing tumor prediction and classification models using methodology and technology of information science based on the tumor gene expression data is meaningful to the research of tumor gene expression patterns identification and tumor diagnosis and recognition as well. This paper presented a method to construct tumor classifier using the classifying rules directly mined from tumor gene expression data. According to this method, we extracted the experiment sample dataset and then searched classifying features that could respectively mark the tumor and normal sample from this dataset. Based on the classifying features mined, the classifying rules were generated and used to predict each unknown sample according to the principle of highest confidence. The experiment made on the prostate cancer gene expression data from Broad Institute showed that the prediction accuracy of this method was over 90 % and a lot of classifying rules with transparent prediction structure were generated at the same time. The experimental results proved the feasibility and effectiveness of this method.

关 键 词：肿瘤 基因表达 数据挖掘 特征选取 分类规则 

分 类 号：R318[医药卫生—生物医学工程]