肝细胞癌（HCC）患者的早期检测和有效的预后预测为改善生存提供了途径，但仍急需更有效的方法。我们寻求基于循环游离线粒体 DNA (ccf-mtDNA) 的片段组学特征，开发超灵敏且低成本的检测和预后模型。对 1168 名参与者的血浆游离 DNA 样本进行了基于捕获的 mtDNA 测序，包括571例HCC患者、301例慢性乙型肝炎或肝硬化(CHB/LC)患者和296名健康对照(HC)。系统分析显示HCC组ccf-mtDNA片段组特征与CHB/LC和HCB/LC组相比显着异常。 HC 组。此外，我们利用ccf-mtDNA片段组特征构建了基于随机森林算法的HCC检测模型。内部和两个外部验证队列均证明我们的模型在区分早期 HCC 患者与 HC 和 CHB/LC 高危人群方面具有出色的能力，AUC 超过 0.983 和 0.981，敏感性超过 89.6% 和 89.61%，特异性超过 98.20 %和95.00%，分别大大超过了甲胎蛋白（AFP）和mtDNA拷贝数的表现。我们还通过 LASSO-Cox 回归开发了 HCC 预后预测模型，以选择 20 个片段组特征，该模型在预测 1 年、2 年和 3 年生存方面表现出卓越的能力（验证队列的 AUC 分别 = 0.8333、0.8145 和 0.7958） ）。我们基于异常的 ccf-mtDNA 片段组特征，在大型临床队列中开发并验证了一种高性能、低成本的方法，在 HCC 患者的早期检测和预后预测方面具有良好的临床转化应用前景。

Early detection and effective prognosis prediction in patients with hepatocellular carcinoma (HCC) provides an avenue for survival improvement, yet more effective approaches are greatly needed. We sought to develop the detection and prognosis models with ultra-sensitivity and low cost based on fragmentomic features of circulating cell free mtDNA (ccf-mtDNA).Capture-based mtDNA sequencing was carried out in plasma cell-free DNA samples from 1168 participants, including 571 patients with HCC, 301 patients with chronic hepatitis B or liver cirrhosis (CHB/LC) and 296 healthy controls (HC).The systematic analysis revealed significantly aberrant fragmentomic features of ccf-mtDNA in HCC group when compared with CHB/LC and HC groups. Moreover, we constructed a random forest algorithm-based HCC detection model by utilizing ccf-mtDNA fragmentomic features. Both internal and two external validation cohorts demonstrated the excellent capacity of our model in distinguishing early HCC patients from HC and high-risk population with CHB/LC, with AUC exceeding 0.983 and 0.981, sensitivity over 89.6% and 89.61%, and specificity over 98.20% and 95.00%, respectively, greatly surpassing the performance of alpha-fetoprotein (AFP) and mtDNA copy number. We also developed a HCC prognosis prediction model by LASSO-Cox regression to select 20 fragmentomic features, which exhibited exceptional ability in predicting 1-year, 2-year and 3-year survival (AUC = 0.8333, 0.8145 and 0.7958 for validation cohort, respectively).We have developed and validated a high-performing and low-cost approach in a large clinical cohort based on aberrant ccf-mtDNA fragmentomic features with promising clinical translational application for the early detection and prognosis prediction of HCC patients.