EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Consensus Integration of Multiomics Data With Machine Learning Algorithms Reveals Heterogeneous Molecular Subtypes and Enables Personalized Treatment Strategies for Hepatocellular Carcinoma

Zhipeng Jin iD, Kun Fang iD, Xue Zhang, Mengying Song, Hong Jiang iD, Yefu Liu iD

Human Mutation · 2025

DOI: 10.1155/humu/9967779 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Human Mutation, 2025. Verfasst von Zhipeng Jin und 5 weiteren Autor:innen. Beteiligte Einrichtung: Department of Hepatopancreatobiliary Surgery, Liaoning Cancer Hospital & Institute, Shenyang, Liaoning, China. Hauptthemen laut MEDLINE: Carcinoma, Hepatocellular, Liver Neoplasms, Machine Learning, Precision Medicine. Die Arbeit wurde bisher 4-mal zitiert (Crossref) und nennt 89 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Background

Cancers are characterized by high heterogeneity. This study seeks to identify the factors driving hepatocellular carcinoma (HCC) heterogeneity to aid in prognostic stratification and inform personalized treatment approaches.

Methods

We used a computational pipeline to integrate multiomics data from HCC patients, applying 10 clustering algorithms. These results were combined with a machine learning framework to identify high‐resolution molecular subtypes (MSs) and to create a robust molecular subtype–related risk score (MSRRS). Subsequent integrated bioinformatics algorithms further analyzed the heterogeneity of HCC at the level of molecular pathways, therapeutic response, and tumor microenvironment, thereby assessing potential clinical value.

Results

Through multiomics clustering, we identified two heterogeneous MSs associated with prognosis, with MS2 exhibiting a more favorable prognostic outcome. Subsequently, we applied bootstrap resampling‐based univariate Cox regression and Boruta algorithm to screen for more clinically relevant genes from the marker genes of each MS. Next, we benchmarked seven survival‐related machine learning algorithms for overall survival (OS) using nested cross‐validation. The hyperparameter‐tuned Ridge survival model outperforms other tuned models and was therefore used to develop a robust MSRRS. MSRRS demonstrated superior performance in predicting patient OS in multiple independent HCC cohorts. Downstream analysis suggested that MSRRS has the potential to guide individualized targeted therapy, chemotherapy, and immunotherapy for HCC and to assess the tumor microenvironment. Pathway enrichment analysis identified the cell cycle as a crucial driver of heterogeneity differences between the two subtypes. Finally, we confirmed that KIF2C may be the most central MSRRS gene and demonstrated by in vitro experiments that KIF2C could promote G2/M transition in HCC cells by targeting CDK1/CCNB1/PLK1 signaling.

Conclusion

The novel MSs and robust MSRRS we identified effectively exposed the heterogeneity of HCC and have the potential to predict prognosis and guide individualized precision therapy.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Carcinoma, HepatocellularLiver NeoplasmsMachine LearningPrecision MedicineHumansPrognosisComputational BiologyAlgorithmsTumor MicroenvironmentBiomarkers, TumorGenomicsMultiomics

Stichwörter der Autor:innen

hepatocellular carcinomaimmunotherapymachine learningmolecular subtypemultiomicsprognosistumor microenvironment

Substanzen

Biomarkers, Tumor

Autor:innen und Institutionen

  • Department of Hepatopancreatobiliary Surgery, Liaoning Cancer Hospital & Institute, Shenyang, Liaoning, China.
  • Central Laboratory, Liaoning Cancer Hospital & Institute, Shenyang, Liaoning, China.
  • Department of Operation Room, Liaoning Cancer Hospital & Institute, Shenyang, Liaoning, China.

ORCID-Profile: Zhipeng Jin, Kun Fang, Hong Jiang, Yefu Liu

Zitieren

APA

Jin, Z., Fang, K., Zhang, X., Song, M., Jiang, H., & Liu, Y. (2025). Consensus Integration of Multiomics Data With Machine Learning Algorithms Reveals Heterogeneous Molecular Subtypes and Enables Personalized Treatment Strategies for Hepatocellular Carcinoma. Human Mutation. https://doi.org/10.1155/humu/9967779

Vancouver

Jin Z, Fang K, Zhang X, Song M, Jiang H, Liu Y. Consensus Integration of Multiomics Data With Machine Learning Algorithms Reveals Heterogeneous Molecular Subtypes and Enables Personalized Treatment Strategies for Hepatocellular Carcinoma. Human Mutation. 2025. doi:10.1155/humu/9967779

BibTeX
@article{Jin2025,
  author = {Jin, Zhipeng and Fang, Kun and Zhang, Xue and Song, Mengying and Jiang, Hong and Liu, Yefu},
  title = {Consensus Integration of Multiomics Data With Machine Learning Algorithms Reveals Heterogeneous Molecular Subtypes and Enables Personalized Treatment Strategies for Hepatocellular Carcinoma},
  journal = {Human Mutation},
  year = {2025},
  doi = {10.1155/humu/9967779}
}
Kostenlos starten →