EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Dynamic differential network construction involving multiple forms of molecular relationships for machine learning-based early clinical diagnosis of hepatocellular carcinoma

Xin Huang, Mei Yan, Xinyu He

Intelligent Medicine · 2026-11

DOI: 10.1016/j.imed.2026.11.006 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Intelligent Medicine, 2026-11. Verfasst von Xin Huang und 2 weiteren Autor:innen. Die Arbeit nennt 32 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: China Postdoctoral Science Foundation, Liaoning Provincial Natural Science Foundation.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Background

Hepatocellular carcinoma (HCC) has a high mortality rate and poses a serious threat to human health. HCC is generally diagnosed at an intermediate or advanced stage. Early and accurate diagnosis of HCC can significantly improve clinical treatment outcomes. A detailed investigation of changes in metabolic activity may improve our understanding of the mechanisms underlying hepatocarcinogenesis and improve early diagnostic performance. However, relationships among metabolism-related molecules, including genes and metabolites, are complex and take different forms.

Methods

A novel dynamic differential network construction algorithm (DDNC) that integrates different forms of molecular relationships is proposed for machine learning-based early diagnosis of HCC. DDNC measures changes in positive linear, negative linear, and nonlinear molecular relationships using a joint probability density function, to reduce the influence of between-sample variability. Experiments used discovery and validation sets of HCC-related genomic and metabolomic data. Performance was evaluated by receiver operating characteristic (ROC) curve analysis and compared with that of five other methods.

Results

For the discrimination of HCC from non-HCC, the area under the ROC curve, standard error, sensitivity, specificity, and 95% confidence interval for the metabolite biomarker panel selected by DDNC were 0.951, 0.035, 0.952, 0.929, and (0.881–1.000) in the metabolomics discovery set, respectively. DDNC also achieved the best performance in terms of all evaluation indices in the metabolomics validation set. DDNC also outperformed the other methods on the genomic datasets in most comparisons. Additionally, partial least squares-discriminant analysis and t-test results further suggested that the biomarker panels selected by DDNC have potential clinical utility for HCC.

Conclusion

By examining changes in multiple forms of molecular relationships rather than linear or nonlinear relationships alone, DDNC integrates these relationships in omics data analysis, potentially improving our understanding of the mechanisms underlying pathological changes.

Förderung

  • China Postdoctoral Science Foundation
  • Liaoning Provincial Natural Science Foundation

Zitieren

APA

Huang, X., Yan, M., & He, X. (2026). Dynamic differential network construction involving multiple forms of molecular relationships for machine learning-based early clinical diagnosis of hepatocellular carcinoma. Intelligent Medicine. https://doi.org/10.1016/j.imed.2026.11.006

Vancouver

Huang X, Yan M, He X. Dynamic differential network construction involving multiple forms of molecular relationships for machine learning-based early clinical diagnosis of hepatocellular carcinoma. Intelligent Medicine. 2026. doi:10.1016/j.imed.2026.11.006

BibTeX
@article{Huang2026,
  author = {Huang, Xin and Yan, Mei and He, Xinyu},
  title = {Dynamic differential network construction involving multiple forms of molecular relationships for machine learning-based early clinical diagnosis of hepatocellular carcinoma},
  journal = {Intelligent Medicine},
  year = {2026},
  doi = {10.1016/j.imed.2026.11.006}
}
Kostenlos starten →