基于机器学习算法筛选鼻咽癌诊断基因标志物的研究  

Screening of genetic markers for diagnosis of nasopharyngeal carcinoma based on machine learning algorithm

作　　者：王艺任 刘艾艾 詹翔 罗颜 周平 WANG Yiren;LIU Aiai;ZHAN Xiang;LUO Yan;ZHOU Ping(School of Nursing,Southwest Medical University,Luzhou,Sichuan,646000;Department of Radiology,the Affiliated Hospital of Southwest Medical University,Luzhou,Sichuan,646000)

机构地区：[1]西南医科大学护理学院,四川泸州646000 [2]西南医科大学附属医院放射科,四川泸州646000

出　　处：《实用临床医药杂志》2023年第7期6-11,共6页Journal of Clinical Medicine in Practice

基　　金：四川省科技计划联合创新专项资助项目(2022YFS0616);四川省医学科研课题计划(S21004);四川省泸州市人民政府-西南医科大学科技战略合作项目(2020LZXNYDJ2);西南医科大学创新创业训练计划(S202210632248)。

摘　　要：目的基于最小绝对收缩和选择算子(LASSO)算法与支持向量机递归特征消除(SVM-RFE)算法筛选用于鼻咽癌(NPC)诊断的特征基因标志物。方法从GEO数据库下载基因表达微阵列数据集GSE53819、GSE13597作为训练集,从GTEx数据库、ICGC数据库分别下载转录组测序数据集GTEx-NPC、ICGC-NPC作为训练集、验证集。通过基因表达差异分析筛选NPC相关差异表达基因(DEGs),再通过LASSO算法和SVM-RFE算法分别筛选3个训练集中的NPC诊断特征基因。结合外部验证集,通过受试者工作特征(ROC)曲线的曲线下面积(AUC)评估特征基因对NPC的诊断效能。结果本研究共筛选出582个NPC相关DEGs,包括156个高表达DEGs和426个低表达DEGs;基于LASSO算法与SVM-RFE算法,GSE53819、GSE13597、GTEx-NPC数据集均筛选出3个关键诊断特征基因HOXA10、AFF3、SHISA3,且GTEx-NPC数据集另有1个特征基因PLAU;ROC曲线分析结果显示,特征基因HOXA10、AFF3、SHISA3、PLAU在各数据集中诊断NPC的AUC均大于0.7,具有良好的诊断效能。结论基于LASSO算法和SVM-RFE算法可筛选出4个潜在的NPC诊断特征基因标志物,且外部验证结果显示这些基因标志物在诊断NPC方面具有良好效能,这为NPC的早期诊断和相关基因的分子机制研究提供了有价值的参考。Objective To screen genetic markers for diagnosis of nasopharyngeal carcinoma(NPC)by the Least Absolute Shrinkage and Selection Operator(LASSO)and Support Vector Machine Recursive Feature Elimination(SVM-RFE)algorithms.Methods Microarray data sets including GSE53819 and GSE13597 of gene expression were downloaded from the GEO database,and transcriptome sequencing data sets including GTEx database and ICGC-NPC database were downloaded as training set and verification set.Differentially expressed genes(DEGs)related to NPC were identified through gene expression differential analysis.Subsequently,LASSO regression and SVM-RFE were used to screen diagnostic feature genes for NPC in three data sets.Finally,an external validation set was used to evaluate the predictive performance of these diagnostic genes by the area under the curve(AUC)of the receiver operating characteristic(ROC)curve.Results A total of 582 DEGs related to NPC were identified,including 156 high expression DEGs and 426 low expression DEGs.Three diagnostic feature genes including HOXA10,AFF3 and SHISA3 were identified by LASSO regression algorithm and SVM-RFE algorithm in the microarray data set.Besides,there was another characteristic gene namely PLAU in the GTEx-NPC dataset.ROC curve analysis results showed that the AUC values of characteristic genes such as HOXA10,AFF3,SHISA3 and PLAU in the diagnosis of NPC in all data sets were greater than 0.7,showing good diagnostic efficacy.Conclusion Four potential diagnostic feature gene markers for NPC based on LASSO and SVM-RFE algorithm are identified,and they provide valuable references for the diagnosis of NPC,showing a valuable reference for the early diagnosis of NPC and the study of the molecular mechanism of related genes.

关 键 词：鼻咽癌 基因组学 机器学习 生物信息学 支持向量机递归特征消除 套索回归 

分 类 号：R739.62[医药卫生—肿瘤] R446[医药卫生—临床医学]