CMGL: Confidence-guided Multi-omics Graph Learning for Cancer Subtype Classification

2026-04-27

AI Summary

This study addresses the challenges in multi-omics cancer subtype classification arising from heterogeneous data quality and noise across modalities. Conventional graph-based methods often fail to reliably model patient similarity due to the tight coupling of modality weighting and classification objectives. To overcome this, the authors propose CMGL, a two-stage framework that first estimates sample-wise, modality-specific confidence scores using evidential deep learning and then leverages these frozen confidence scores to guide cross-omics fusion and graph construction, thereby decoupling reliability assessment from graph learning. Notably, CMGL introduces sample-level modality confidence as an independent prior, effectively mitigating the adverse influence of low-quality modalities on graph topology and message passing. Evaluated on four single-cancer tasks, CMGL achieves an average accuracy improvement of 4.03% over the strongest baseline, successfully recapitulates the PAM50 breast cancer subtypes, and demonstrates zero-shot transferability to kidney cancer with significant prognostic stratification.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Computer Vision: Multi-modal Vision
• Reasoning under Uncertainty: Graphical Models

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Semantics and Knowledge: Scalable techniques for the creation, curation, publication, maintenance, and consumption of large, Web-based, structured, reusable, knowledge graphs and ontologies
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data

Abstract

Motivation: Multi-omics integration can improve cancer subtyping, but modality informativeness and noise vary across cancer types and patients. Existing graph-based methods optimize modality weights jointly with the classification objective and therefore lack independent reliability estimates, so low-quality omics distort patient similarity graphs and amplify noise through message passing. Results: We propose CMGL, a two-stage framework that estimates per-sample modality reliability through evidential deep learning and uses the frozen confidence scores to guide cross-omics fusion and graph construction. 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 BRCA-trained model transfers without fine-tuning to kidney renal clear cell carcinoma (KIRC), stratifying patients into prognostically distinct groups.

Problem

• multi-omics integration
• cancer subtype classification
• modality reliability
• patient similarity graph
• noise amplification

Innovation

• confidence-guided learning
• multi-omics integration
• evidential deep learning
• graph neural networks
• cancer subtype classification