目的

探讨循环肿瘤DNA（ctDNA）甲基化预测结直肠癌（CRC）模型构建与验证。

方法

2021 年8 月至2023 年7 月前瞻性收集徐州市中心医院收治的CRC 患者作为研究对象（CRC组），同时期年龄和性别匹配的健康体检者作为对照组。比较两组相关临床资料、血清ctDNA 甲基化水平；通过公共数据库“Marmal-aid”获得CRC 的DNA 甲基化标志物；通过LASSO 和RF 分析筛选DNA 甲基化标志物；通过ROC 曲线和Logistic 多因素回归分析，构建基于筛选的DNA 甲基化标志物预测CRC 的模型；通过训练队列和验证队列分别明确本模型对CRC 的预测效能。

结果

通过公共数据库进行筛选，共鉴定出3997 个差异性DNA 甲基化CpG（DMCs），并通过分层聚类和KEGG 显示出CRC 和正常结直肠黏膜在这些DMCs 中不同的甲基化模式及富集通路，并逐步筛选保留了19 个甲基化标志物。通过LASSO 和RF 分析评估了19 个标志物的稳定性，获得了3 个DNA甲基化标志物（cg08131100、cg16934178 和cg24171907），预测CRC 的AUC 分别为0.887、0.841和0.715。Logistic 回归分析显示，cg08131100、cg16934178 和cg24171907 是影响CRC 发病的独立危险因素；在训练队列和验证队列中，CRC 的甲基化评分均显著高于对照组；通过ROC 曲线评价甲基化模型预测CRC 的效能，模型在训练组AUC 为0.841（95% CI 0.756～0.927），模型在验证组AUC 为0.823（95% CI 0.728～0.919）。

结论

本研究基于ctDNA 甲基化构建了预测CRC 患者的模型，为CRC 的无创检测和筛查提供了具有临床转化意义的工具。

Objective

To investigate the construction and validation of a model for prediction of colorectal cancer （CRC） by circulating tumor DNA （ctDNA） methylation.

Methods

Prospectively enrolled colorectal cancer patients between August 2021 and July 2023 were selected as the study subjects（CRC group），and age-and gender-matched healthy medical check-ups in the same period served as the control group. The two groups were compared in terms of clinical data and serum ctDNA methylation levels； DNA methylation markers of CRC were obtained from the public database “Marmal-aid”； DNA methylation markers were screened by LASSO and RF analyses. The model of CRC prediction based on the screened DNA methylation markers was constructed by ROC curve and logistic multifactorial regression analysis； the prediction efficacy of the model for CRC was clarified by the training cohort and validation cohort，respectively.

Results

A total of 3997 differentially DNA methylated CpGs （DMCs）were identified by screening through public databases，and 19 methylation markers were retained by stepwise screening through hierarchical clustering and KEGG showing the different methylation patterns and enrichment pathways between CRC and normal colorectal mucosa in these DMCs. The stability of the 19 markers was assessed by LASSO and RF analyses，and three DNA methylation markers （cg08131100，cg16934178，and cg24171907） were obtained，with predicted AUCs of 0.887，0.841，and 0.715 for CRC，respectively. Logistic regression analyses showed that cg08131100，cg16934178 and cg24171907 were independent risk factors influencing the onset of CRC； the methylation scores of CRC were significantly higher than those of the control group in both the training and validation cohorts； and the efficacy of the methylation model in predicting CRC was evaluated by ROC curves，and the model had an AUC of 0.841 in the training group （95% CI： 0.756-0.927），and the AUC of the model in the validation group was 0.823 （95%CI： 0.728-0.919）.

Conclusion

This study constructs a model for prediction of CRC patients based on ctDNA methylation，which provides a clinically translatable tool for noninvasive detection and screening of CRC.