利用基因表达值相对大小秩序标志鉴别肺癌

发布时间：2018-04-23 21:33

本文选题：标志 + 分类器　；参考：《生物医学工程学杂志》2017年01期

【摘要】：在应用基于转录组特征构建的支持向量机、贝叶斯分类器等传统分类器对组织样本进行分类时,要求对基因表达谱进行样本间的数据标准化处理,以去除实验批次效应带来的影响,因此限制了这些分类器在个体化水平上的应用。本文旨在构建鉴别肺癌组织与非癌(肺炎与肺正常)组织的个体化分类器。文中采用来自多组独立数据的197例肺癌与189例肺非癌组织样本作为训练集,筛选得到了3对基因作为特征,应用多数投票规则区分肺癌组织与肺非癌组织的平均准确率达到95.34%。然后,本文采用来自多组独立数据的251例肺癌组织与141例肺非癌组织样本的非标化数据进行独立验证,其平均准确率达到96.78%。因此,本文提出的该分类器可对由不同实验室检测的样本进行个体化判断提供一种新的思路,具有较强的临床实用性。
[Abstract]:When traditional classifiers such as support vector machine and Bayesian classifier based on transcriptional group feature are used to classify tissue samples, it is required to standardize the data of gene expression profiles among samples. The application of these classifiers at the individualized level is limited by removing the effects of the experimental batch effect. The purpose of this study was to construct an individual classifier for distinguishing lung cancer from non-cancerous (pneumonia and normal lung) tissues. In this paper, 197 samples of lung cancer and 189 samples of non-cancerous lung tissue from multiple independent data were used as training set, and 3 pairs of genes were selected as characteristics. The average accuracy of majority voting rule in distinguishing lung cancer tissue from non-cancerous lung tissue was 95.3434. Then, the nonstandard data from 251 lung cancer tissues and 141 noncancerous lung tissue samples from multiple groups of independent data were used for independent verification. The average accuracy was 96.78%. Therefore, the classifier presented in this paper can provide a new idea for individualized judgment of samples detected by different laboratories, and it has strong clinical practicability.
【作者单位】：福建医科大学基础医学院生物信息学系消化道恶性肿瘤教育部重点实验室;
【基金】：国家自然科学基金(81572935,81372213,81501215,81501829) 大学生创新创业训练计划(201510392017) 福建医科大学苗圃科研基金(2015MP005)
【分类号】：R734.2

【相似文献】