基于TCGA和ICGC数据库的线粒体自噬相关基因（MRGs）构建肝细胞癌（肝癌）患者预后风险模型，并筛选具有治疗潜力的小分子药物。

从TCGA和ICGC数据库中下载肝癌患者的RNA测序数据和临床信息。从Reactome信号通路数据库检索收集MRGs。采用单因素Cox回归和Lasso回归分析以TCGA队列为训练集构建预后风险模型，并在ICGC队列中进行验证。R软件"rms"包用于构建列线图。R软件"limma"包用于选择预后风险模型中高风险组和低风险组的差异基因，利用获得的差异基因在联系图（CMap）数据库中筛选具有治疗潜力的小分子药物。

19个MRGs在TCGA和ICGC队列肿瘤组织中均表达上调，PINK表达下调。12个MRGs在两个队列中均为生存预后的危险因素。利用Lasso回归分析构建5-MRGs标志物的预后风险模型，包括酪蛋白激酶2β多肽（CSNK2B）、线粒体融合蛋白1（MFN1）、磷酸甘油酸突变酶家族成员5（PGAM5）、外线粒体膜转位酶同源物（TOMM）5、TOMM22基因。Kaplan-Meier生存分析显示，训练集和验证集中高风险组的预后较差，中位生存期较短。ROC曲线提示该模型对肝癌患者预后有较高的预测价值。多因素Cox分析提示线粒体自噬评分是影响肝癌患者预后的独立因素（HR=2.68，95%CI：1.65~4.36，P<0.001）。在TCGA队列中结合肿瘤分期、分级、T分期与线粒体自噬评分构建了列线图预测患者的生存率。从CMap数据库中筛选出4个有潜力可逆转高风险组预后较差的特征小分子药物，分别为DL-thiorphan、blebbistatin、talampicillin和puromycin。

本研究基于MRGs构建了一个稳定的预后风险模型和列线图，并通过不同风险组的差异基因筛选出可能具有治疗作用的小分子药物。

To establish a prognostic risk model for hepatocellular carcinoma (HCC) patients based on mitophagy-related genes (MRGs) from TCGA and ICGC databases, and to screen small molecule drugs with therapeutic potential.

RNA sequencing data and clinical information of HCC patients were downloaded from TCGA and ICGC databases. MRGs were retrieved and collected from Reactome signaling pathway database. Univariate Cox regression and Lasso regression analyses were employed to establish the prognostic risk model using TCGA cohort as the training data set and ICGC cohort as the verification set. The nomogram was plotted with R software "rms" package. The differential genes between the high-risk and low-risk groups in the prognostic risk model were selected with R software "limma" package. Small molecule drugs with therapeutic potential were screened out from the CMap database using the differential genes.

The expression levels of 19 MRGs were up-regulated, whereas that of PINK was down-regulated in the tumor tissues of TCGA and ICGC cohorts. All 12 MRGs were the risk factors for the clinical prognosis in two cohorts. A prognostic risk model of 5-MRGs markers was established by Lasso regression analysis, including casein kinase 2β polypeptide (CSNK2B), mitofusin 1 (MFN1), phosphoglycerate mutase family member 5 (PGAM5), translocase of outer mitochondrial membrane (TOMM) 5 and TOMM22 genes. Kaplan-Meier survival analysis demonstrated that prognosis in the high-risk groups in training and verification sets was poorer, and the median survival was shorter. ROC curve indicated this model yielded high predictive value for the prognosis of HCC patients. Multivariate Cox analysis demonstrated that mitophagy score was an independent factor affecting the prognosis of HCC patients (HR=2.68, 95%CI: 1.65-4.36, P<0.001). In the TCGA cohort, a nomogram was plotted combining tumor staging, grading, T staging and mitophagy score to predict the survival of HCC patients. 4 small molecular drugs with the potential to reverse the poor prognosis of patients in high-risk groups were screened out from the CMap database, including DL-thiorphan, blebbistatin, talampicillin and puromycin.

In this study, a stable prognostic risk model and nomogram are constructed based on MRGs, and small molecular drugs that might have therapeutic potential are screened out by differential genes in different risk groups.