Related Experiment Video
Updated: Jul 12, 2026
A Hepatocellular Cancer Patient-Derived Organoid Xenograft Model to Investigate Impact of Liver Regeneration on Tumor Growth
Published on: February 2, 2024
Identification of therapeutic targets and postoperative recurrence prediction model construction for hepatocellular
Yang Kong1, Yizhi Wang1, Qifan Yang1
1Department of Hepatobiliary and Pancreatic Surgery, The Second Affiliated Hospital, School of Medicine, Zhejiang University, Hangzhou, China; Key Laboratory, Precision Diagnosis and Treatment for Hepatobiliary and Pancreatic Tumor of Zhejiang Province, Hangzhou, China; Research Center, Diagnosis and Treatment Technology for Hepatocellular Carcinoma of Zhejiang Province, Hangzhou, China; Clinical Medicine Innovation Center, Precision Diagnosis and Treatment for Hepatobiliary and Pancreatic Disease, Zhejiang University, Hangzhou, China; Clinical Research Center of Hepatobiliary and Pancreatic Diseases of Zhejiang Province, Hangzhou, China; Cancer Center, Zhejiang University, Hangzhou, China.
Background And Study Aims:
Hepatocellular carcinoma (HCC) has a high risk of postoperative recurrence. This study aimed to develop a transcriptome-based recurrence prediction model and explore molecular features using Mendelian randomization (MR), immune infiltration, and single-cell RNA sequencing.
Materials And Methods:
Transcriptomic data from TCGA-LIHC (n = 415), GSE14520 (n = 240), and GSE16757 (n = 101) were integrated to identify differentially expressed genes (DEGs). A total of 113 machine-learning strategies, including standalone models and feature-selection/classifier combinations, were evaluated, and the RF model was selected according to average AUC performance. Bootstrap optimism correction and exploratory external validation using GSE164368 were performed to assess model robustness. MR analysis was used to evaluate the genetic association between COL2A1 and HCC risk, while immune infiltration and single-cell RNA-seq analyses were conducted to explore its tumor microenvironmental relevance.
Results:
A total of 141 DEGs associated with HCC recurrence were identified, with RF highlighting 24 core genes. The RF model achieved the highest average AUC of 0.930, retained stable performance after bootstrap optimism correction, and showed exploratory external validation performance in GSE164368 with an AUC of 0.889. MR analysis showed that increased COL2A1 expression was significantly protective against HCC (OR = 0.811; 95% CI, 0.669-0.984; P = 0.034). Single-cell RNA-seq analysis showed higher COL2A1 expression in hepatocytes, macrophages, endothelial cells, and tissue stem cells from MVI-negative samples than in MVI-positive samples. Functional enrichment and immune infiltration analyses suggested that COL2A1 may be associated with recurrence- and invasion-related biological processes, including extracellular matrix remodeling, epithelial-mesenchymal transition (EMT), and immune microenvironment regulation.
Conclusion:
The RF-based 24-gene model showed promising performance for classifying postoperative recurrence in HCC. COL2A1 may be linked to recurrence-associated microenvironmental states, involving extracellular matrix remodeling and immune regulation. Further validation is required before clinical translation.
