aiqbench analysis
CMGL introduces a novel graph learning framework that explicitly quantifies modality reliability using evidential deep learning before fusing multi-omics data for cancer subtyping. By decoupling confidence estimation from the classification objective, the method effectively mitigates noise propagation and demonstrates strong cross-cancer transferability without fine-tuning. This approach advances the state of multi-modal integration in genomics by providing more robust representations for heterogeneous patient populations.
CMGL: Confidence-guided Multi-omics Graph Learning for Cancer Subtype Classification
arXiv:2604.24201v2 Announce Type: replace 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 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 (KIRC), stratifying patients into prognostically distinct groups.
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 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 (KIRC), stratifying patients into prognostically distinct groups.
Machine Learning (cs.LG); Genomics (q-bio.GN); Molecular Networks (q-bio.MN)
arXiv:2604.24201 [cs.LG]
(or arXiv:2604.24201v2 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2604.24201 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 273 words
Related coverage
Beyond SBDD: Geometric Deep Learning in Polypharmacology and Multi-target Drug Design
arXiv · Jul 24, 2026
Evaluating Single and Multi-Omics Based Explainable Artificial Intelligence (MOXAI) for Molecular Subclass Classification of Adult-Type Diffuse Gliomas
arXiv · Sep 29, 2026
DoGMA: A Central-Dogma-Guided Foundation Model for Multi-Omics Alignment and Multi-Task Learning in Oncology
arXiv · Sep 9, 2026
Diagnosing Multi-step Reasoning Failures in Black-box LLMs via Stepwise Confidence Attribution
arXiv · Jun 9, 2026