基于H4C6甲基化水平和cfDNA浓度构建多癌种癌症风险预测模型

A multi-cancer risk prediction model which constructed based on H4C6 methylation level and cfDNA concentration

目的 探究H4聚簇组蛋白6(H4C6)甲基化水平和循环游离DNA(cfDNA)浓度在正常组与肿瘤组之间的差异;基于H4C6甲基化水平和cfDNA浓度构建癌症风险预测模型并评估模型的预测性能。方法 使用磁珠法提取血液样本中的cfDNA;使用Qubit 4.0荧光定量仪检测cfDNA浓度;利用实时荧光定量PCR(RT-qPCR)技术检测cfDNA的H4C6甲基化水平;使用Logistic回归算法构建H4C6甲基化水平联合cfDNA浓度的癌症风险预测模型;使用受试者工作特征(ROC)曲线和校准曲线评估模型的准确性;使用决策曲线分析(DCA)评估模型的临床效益。结果 联合H4C6甲基化水平与cfDNA浓度构建的模型区分肺癌、肝癌、结直肠癌、胃癌、泛癌与健康组的ROC曲线下面积(AUC)分别为0.769、0.988、0.934、0.922、0.830;校准曲线的平均绝对误差小于0.05;DCA曲线的净收益大于0。结论 基于H4C6甲基化水平和cfDNA浓度构建的癌症风险预测模型有较好的预测性能,有助于给临床前决策提供合理且有效的建议,最终可能给患者提供有针对性的、个性化的癌症检测与诊断方案。

Objective To explore the difference in H4 clustered histone 6(H4C6)methylation level and circulating cell-free DNA(cfDNA)concentration between 94 normal group and 122 tumor groups(65 patients with lung cancer,22 patients with gastric cancer,23 patients with colorectal cancer,and 12 patients with liver cancer),and the age of total 216 subjects were between 18 and 85 years old.To construct a cancer risk prediction model based on H4C6 methylation level and cfDNA concentration and evaluate the predictive performance of the model.Methods cfDNA was extracted from blood samples using magnetic beads.Qubit 4.0 fluorescence quantitative meter was used to detect the concentration of cfDNA.Real-time quantitative PCR(RT-qPCR)technology was used to detect the methylation level of H4C6 in cfDNA.Logistic regression algorithm was used to construct a cancer risk prediction model of H4C6 methylation level combined with cfDNA concentration.The accuracy of the model was assessed using receiver operating characteristic(ROC)curve and calibration curve.The clinical benefit of the model was assessed using decision curve analysis(DCA).Results The model was constructed by combining H4C6 methylation level and cfDNA concentration to distinguish lung cancer,liver cancer,colorectal cancer,gastric cancer,pancancer from healthy control group had the area under curve(AUC)of 0.769,0.988,0.934,0.922,0.830,respectively.The mean absolute error of the calibration curve was less than 0.05;the net benefit of the DCA curve was greater than 0.Conclusion The cancer risk prediction model based on H4C6 methylation level and cfDNA concentration has good predictive performance,which helps to provide reasonable and effective suggestions for preclinical decision-making,and ultimately may provide patients with targeted and personalized cancer detection and diagnosis program.