基于癫痫患者单核苷酸多态性的丙戊酸钠疗效预测模型

Prediction Model of Valproate Treatment Efficacy in Epileptic Patients Based on Single Nucleotide Polymorphism

目的:构建基于单核苷酸多态性(SNP)的癫痫患者丙戊酸钠(VPA)疗效预测模型.方法:对服用VPA一年以上的癫痫患者进行基因分型,利用随机森林筛选与疗效相关的SNP,分别通过Logistics回归模型和随机森林模型构建VPA疗效预测模型,采用五折交叉验证法对两种预测模型进行验证,并比较两种模型的预测效果.结果:共有122名符合纳入排除标准的癫痫患者,平均年龄为(27.20±14.57)岁,男性占比为67.21%.Logistics预测模型的AUC及95%CI为0.825(0.722,0.928),灵敏度为77.6%,特异度为80.0%;随机森林预测模型其AUC及95%CI为0.738(0.750,0.961),灵敏度为63.6%,特异度为80.0%.Logistic预测模型结果优于随机森林预测模型,模型预测因子为rs4646440、rs4984241、rs2894342、rs12233719、rs1019385这5个SNP,并根据模型结果绘制了列线图(Nomogram).结论:本研究构建了基于单核苷酸多态性的癫痫患者丙戊酸钠疗效预测模型,为癫痫患者个性化治疗提供了一定的科学参考.

Objective:To establish a prediction model of Valproate(VPA)treatment efficacy based on single nucleotide polymorphism(SNP).Methods:Genotyping for epileptic patients with VPA who had been treated for more than one year was conducted.SNPs were screened by random forests.VPA efficacy prediction models were constructed by Logistics regression model and random forest model,respectively.The five-fold cross-validation method was used to verify the two prediction models,and the prediction effects of the two models were compared.Results:A total of 122 epilepsy patients met the inclusion and exclusion criteria,with an average age of(27.20±14.57)years old and a male proportion of 67.21%.The AUC and 95%CI of Logistics prediction model were 0.825(0.722,0.928),the sensitivity was 81.3%,and the specificity was 80.0%.The AUC and 95%CI of the random forest prediction model were 0.738(0.750,0.961),the sensitivity was 63.6%,and the specificity was 80.0%.The results of the Logistic prediction model were superior to those of the random forest prediction model.The prediction factors of the model were rs4646440,rs4984241,rs2894342,rs12233719 and rs1019385.Conclusion:A single nucleotide polymorphism-based prediction model for the VPA efficacy in epilepsy patients was established,which provides a certain scientific reference for personalized treatment of epilepsy patients.