School of Public Health, Wenzhou Medical University, Wenzhou, China.
Clin Exp Pharmacol Physiol. 2024 Aug;51(8):e13907. doi: 10.1111/1440-1681.13907.
Most cases of hepatocellular carcinoma (HCC) arise as a consequence of cirrhosis. In this study, our objective is to construct a comprehensive diagnostic model that investigates the diagnostic markers distinguishing between cirrhosis and HCC.
Based on multiple GEO datasets containing cirrhosis and HCC samples, we used lasso regression, random forest (RF)-recursive feature elimination (RFE) and receiver operator characteristic analysis to screen for characteristic genes. Subsequently, we integrated these genes into a multivariable logistic regression model and validated the linear prediction scores in both training and validation cohorts. The ssGSEA algorithm was used to estimate the fraction of infiltrating immune cells in the samples. Finally, molecular typing for patients with cirrhosis was performed using the CCP algorithm.
The study identified 137 differentially expressed genes (DEGs) and selected five significant genes (CXCL14, CAP2, FCN2, CCBE1 and UBE2C) to construct a diagnostic model. In both the training and validation cohorts, the model exhibited an area under the curve (AUC) greater than 0.9 and a kappa value of approximately 0.9. Additionally, the calibration curve demonstrated excellent concordance between observed and predicted incidence rates. Comparatively, HCC displayed overall downregulation of infiltrating immune cells compared to cirrhosis. Notably, CCBE1 showed strong correlations with the tumour immune microenvironment as well as genes associated with cell death and cellular ageing processes. Furthermore, cirrhosis subtypes with high linear predictive scores were enriched in multiple cancer-related pathways.
In conclusion, we successfully identified diagnostic markers distinguishing between cirrhosis and hepatocellular carcinoma and developed a novel diagnostic model for discriminating the two conditions. CCBE1 might exert a pivotal role in regulating the tumour microenvironment, cell death and senescence.
大多数肝细胞癌 (HCC) 是肝硬化的后果。本研究旨在构建一个综合的诊断模型，以研究区分肝硬化和 HCC 的诊断标志物。
基于包含肝硬化和 HCC 样本的多个 GEO 数据集，我们使用lasso 回归、随机森林 (RF)-递归特征消除 (RFE) 和接收者操作特征分析筛选特征基因。随后，我们将这些基因整合到多变量逻辑回归模型中，并在训练和验证队列中验证线性预测分数。使用 ssGSEA 算法估计样本中浸润免疫细胞的分数。最后，使用 CCP 算法对肝硬化患者进行分子分型。
研究确定了 137 个差异表达基因 (DEGs)，并选择了五个显著基因 (CXCL14、CAP2、FCN2、CCBE1 和 UBE2C) 构建诊断模型。在训练和验证队列中，该模型的 AUC 均大于 0.9，kappa 值约为 0.9。此外，校准曲线显示观察到的和预测的发病率之间具有极好的一致性。相比之下，与肝硬化相比，HCC 显示浸润免疫细胞总体下调。值得注意的是，CCBE1 与肿瘤免疫微环境以及与细胞死亡和细胞衰老过程相关的基因具有很强的相关性。此外，线性预测得分较高的肝硬化亚型在多个癌症相关途径中富集。
总之，我们成功地鉴定了区分肝硬化和肝细胞癌的诊断标志物，并开发了一种用于区分两种情况的新诊断模型。CCBE1 可能在调节肿瘤微环境、细胞死亡和衰老中发挥关键作用。
