Skip to content
DraconiticResearch
Search Sources API
197,768 studies31 sourcesUpdated 23 min ago
Peer-reviewedCC BY 4.0Bioinformatics

CMGL: confidence-guided multi-omics graph learning for cancer subtype classification

  • Boyang Fan1
  • Hengchuang Yin2
  • Siyu Yi3
  • Yifan Wang4
  • Zhicheng Li1
  • Leijiyu Zhou1
  • Jiancheng Lv1
  • Wei Ju5
5 affiliations
  1. 1College of Computer Science, Sichuan University , Chengdu 610065,
  2. 2Xinjiang Technical Institute of Physics and Chemistry, Chinese Academy of Sciences , Urumqi 830011,
  3. 3College of Mathematics, Sichuan University , Chengdu 610065,
  4. 4School of Artificial Intelligence and Data Science, University of International Business and Economics , Beijing 100029,
  5. 5School of Artificial Intelligence, Sichuan University , Chengdu 610065,
Read full textPDFDOICite

Abstract

Motivation: Multi-omics integration can improve cancer subtyping, but modality informativeness and noise vary across cancer types and patients. Most graph methods for multi-omics data learn modality contributions within the downstream classification objective, leaving predictive reliability for each patient implicit. As a result, uninformative modalities can weaken the fused representation, while unreliable omics can introduce noisy patient relationships into graph propagation. To address these two problems, we propose confidence-guided multi-omics graph learning (CMGL), which produces a separate reliability estimate before fusion and uses consensus patient neighborhoods for graph classification.

Results: CMGL estimates modality confidence for each patient through evidential deep learning, fixes these values during fusion across omics, and performs classification on an independently specified consistency graph. On four MLOmics cancer-subtype tasks and the 32-class pan-cancer task, CMGL consistently improves over the strongest baseline, surpassing it by 4.03% in average accuracy on the four single-cancer tasks. Its representations recover the PAM50 intrinsic subtypes of breast invasive carcinoma (BRCA), and the model trained on BRCA transfers without fine tuning to kidney renal clear cell carcinoma, stratifying patients into prognostically distinct groups.

Availability and implementation: The code is available at https://doi.org/10.5281/zenodo.21921069.

Related

  1. GONet: A Molecular Interaction-Guided Graph Learning Framework for Multi-Omics Cancer Classification
    Yiyang Chen, Yunning Lu, Da Qi +3
    GigaSciencePeer-reviewedCC BY 4.0Full textMolecular biology & gene regulationCrossref
  2. Explainable computational framework for subtype classification and biomarker identification in head and neck squamous cell carcinoma using machine learning
    Anagha Rajan, Christopher Leela Biji, Mangalam Goutam Sinha +1
    PLOS OnePeer-reviewedCC BY 4.0Full textMolecular biology & gene regulationBioinformatics & computational biologyPLOSPMC
  3. Confidence-guided cryo-EM map optimisation with LocScale-2.0
    Alok Bharadwaj, Reinier De Bruin, Arjen J. Jakobi
    Nature CommunicationsPeer-reviewedCC BY13 citedStructural biologyOpenAlex
  4. Early diagnosis and molecular classification of gastric cancer peritoneal metastasis by deep learning-driven single-cell lipidomics
    Siyuan Pan, Xiaowan Chen, Wen Xu +21
    Cell Reports MedicinePeer-reviewedCC BY-NC 4.0Full textMolecular biology & gene regulationPMC
  5. Subtype-Specific Aptamer Technologies for Breast Cancer: From Molecular Recognition to Precision Applications.
    Huibin Wu, Zhi Zeng, Wenyi Huang +4
    BiosensorsPeer-reviewedCC BYFull textMolecular biology & gene regulationEurope PMC

Cite

BibTeXRIS
BibTeX
@article{fan2026cmgl,
  title = {{CMGL: confidence-guided multi-omics graph learning for cancer subtype classification}},
  author = {Boyang Fan and Hengchuang Yin and Siyu Yi and Yifan Wang and Zhicheng Li and Leijiyu Zhou and Jiancheng Lv and Wei Ju},
  journal = {Bioinformatics},
  year = {2026},
  month = oct,
  doi = {10.1093/bioinformatics/btag763},
  url = {https://academic.oup.com/bioinformatics/article/doi/10.1093/bioinformatics/btag763/8911834},
  note = {CC BY 4.0},
}
RIS
TY  - JOUR
TI  - CMGL: confidence-guided multi-omics graph learning for cancer subtype classification
AU  - Boyang Fan
AU  - Hengchuang Yin
AU  - Siyu Yi
AU  - Yifan Wang
AU  - Zhicheng Li
AU  - Leijiyu Zhou
AU  - Jiancheng Lv
AU  - Wei Ju
PY  - 2026
DA  - 2026/10/12
JO  - Bioinformatics
DO  - 10.1093/bioinformatics/btag763
UR  - https://academic.oup.com/bioinformatics/article/doi/10.1093/bioinformatics/btag763/8911834
AB  - Motivation: Multi-omics integration can improve cancer subtyping, but modality informativeness and noise vary across cancer types and patients. Most graph methods for multi-omics data learn modality contributions within the downstream classification objective, leaving predictive reliability for each patient implicit. As a result, uninformative modalities can weaken the fused representation, while unreliable omics can introduce noisy patient relationships into graph propagation. To address these two problems, we propose confidence-guided multi-omics graph learning (CMGL), which produces a separate reliability estimate before fusion and uses consensus patient neighborhoods for graph classification.  Results: CMGL estimates modality confidence for each patient through evidential deep learning, fixes these values during fusion across omics, and performs classification on an independently specified consistency graph. On four MLOmics cancer-subtype tasks and the 32-class pan-cancer task, CMGL consistently improves over the strongest baseline, surpassing it by 4.03% in average accuracy on the four single-cancer tasks. Its representations recover the PAM50 intrinsic subtypes of breast invasive carcinoma (BRCA), and the model trained on BRCA transfers without fine tuning to kidney renal clear cell carcinoma, stratifying patients into prognostically distinct groups.  Availability and implementation: The code is available at https://doi.org/10.5281/zenodo.21921069.
SN  - 1367-4803
SN  - 1367-4811
ER  -