Scholarlink

Purpose <p>Microvascular invasion (MVI) is an independent risk factor for hepatocellular carcinoma (HCC) recurrence and metastasis, but its biomarkers remain unknown. This study aimed to identify multi-omics annotated biomarkers of MVI and to construct a survival prediction model for HCC based on multimodal features.</p> Patients and methods <p>Patients with pathologically confirmed HCC from January 2017 to June 2019 were retrospectively analyzed. Radiomic features were extracted from tumor areas and different peri-tumor areas in preoperative CT images. In the training cohort, three radiomic models were constructed after the dimensionality reduction of these features to predict MVI. The model performance was tested according to the pathological reference standard using the internal test cohort, and the area under the subject operating characteristic curve (AUC) was calculated. The Rad-score of the optimal model was used to assess the predictive value of the radiomic model for overall survival (OS). RNA sequencing data from the Cancer Genome Atlas (TCGA) were analyzed to explore gene expression and the immune microenvironment. The predictors of OS were determined by univariate and multifactorial Cox proportional regression, and a combined model was further constructed to predict OS in HCC patients.</p> Results <p>Of 184 patients, 97 had MVI. The model constructed with tumor combined with the 5mm peritumor region had the best performance in predicting MVI, and the AUC of training cohort and validation cohort were 0.80 and 0.79, respectively. The Rad-score divided patients into high-risk and low-risk groups, with significant differences in survival in the training set and the internal test set and in the TCGA (<i>P</i> &lt; 0.0001, P &lt; 0.0001 and <i>P</i> = 0.019). Five MVI-related genes (SLC44A4, LAMP3, TOX, PTGS1 and MS4A4A) were identified by Rad-score. Univariate and multivariate Cox analyses showed that Rad-score, Gene-score, and Visceral adipose tissue (VAT) were independent risk factors for survival in HCC patients. The combined model achieved a 1-year, 3-year, and 5-year TD AUC for OS of 0.818 and 0.880.0.883, respectively. In addition, Immune microenvironment analysis showed significant differences in macrophages-M0 and IPS scores among different risk groups.</p> Conclusion <p>Multi-modal biomarkers of MVI can be identified based on CT images, and the overall survival of HCC patients can be accurately predicted.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Identifying potential biomarkers of microvascular invasion in hepatocellular carcinoma and predict overall survival based on multi-omics

  • 期刊论文

  • 发表日期：

    2025年12月10日

查看全文
  • Zhongqi Sun,
  • Kai Zhao,
  • Jin Zhang,
  • Qiong Wu,
  • Linhan Zhang,
  • Xue Lin,
  • Yanjie Xin,
  • Jinping Li,
  • Huijie Jiang

摘要

Purpose

Microvascular invasion (MVI) is an independent risk factor for hepatocellular carcinoma (HCC) recurrence and metastasis, but its biomarkers remain unknown. This study aimed to identify multi-omics annotated biomarkers of MVI and to construct a survival prediction model for HCC based on multimodal features.

Patients and methods

Patients with pathologically confirmed HCC from January 2017 to June 2019 were retrospectively analyzed. Radiomic features were extracted from tumor areas and different peri-tumor areas in preoperative CT images. In the training cohort, three radiomic models were constructed after the dimensionality reduction of these features to predict MVI. The model performance was tested according to the pathological reference standard using the internal test cohort, and the area under the subject operating characteristic curve (AUC) was calculated. The Rad-score of the optimal model was used to assess the predictive value of the radiomic model for overall survival (OS). RNA sequencing data from the Cancer Genome Atlas (TCGA) were analyzed to explore gene expression and the immune microenvironment. The predictors of OS were determined by univariate and multifactorial Cox proportional regression, and a combined model was further constructed to predict OS in HCC patients.

Results

Of 184 patients, 97 had MVI. The model constructed with tumor combined with the 5mm peritumor region had the best performance in predicting MVI, and the AUC of training cohort and validation cohort were 0.80 and 0.79, respectively. The Rad-score divided patients into high-risk and low-risk groups, with significant differences in survival in the training set and the internal test set and in the TCGA (P < 0.0001, P < 0.0001 and P = 0.019). Five MVI-related genes (SLC44A4, LAMP3, TOX, PTGS1 and MS4A4A) were identified by Rad-score. Univariate and multivariate Cox analyses showed that Rad-score, Gene-score, and Visceral adipose tissue (VAT) were independent risk factors for survival in HCC patients. The combined model achieved a 1-year, 3-year, and 5-year TD AUC for OS of 0.818 and 0.880.0.883, respectively. In addition, Immune microenvironment analysis showed significant differences in macrophages-M0 and IPS scores among different risk groups.

Conclusion

Multi-modal biomarkers of MVI can be identified based on CT images, and the overall survival of HCC patients can be accurately predicted.