Skip to content
DraconiticResearch
Search Sources API
190,869 studies31 sourcesUpdated 45 min ago
Peer-reviewedCC BY-NCBriefings in bioinformatics

Multi-omics survival prediction under modality incompleteness via dynamic modality graphs and product-of-experts fusion.

  • Weidong Xie1,2,3
  • Huixia Zhang1,2,3
  • Jinzhu Yang1,2,3
  • Kun Yu1,4
  • Wei Li1,2,3
4 affiliations
  1. 1School of Computer Science and Engineering, Northeastern University, 195 Chuangxin Road, Hunnan District, Shenyang, Liaoning 110169, China.
  2. 2Key Laboratory of Intelligent Computing in Medical Image (MIIC), Northeastern University, No. 11, Lane 3, Wenhua Road, Heping District, Shenyang, Liaoning 110819, China.
  3. 3National Frontiers Science Center for Industrial Intelligence and Systems Optimization, Northeastern University, No. 11, Lane 3, Wenhua Road, Heping District, Shenyang, Liaoning 110819, China.
  4. 4College of Medicine and Bioinformation Engineering, Northeastern University, 195 Chuangxin Road, Hunnan District, Shenyang, Liaoning 110169, China.
Read full textPDFDOICite

Abstract

Multi-omics data integration provides rich complementary information for cancer survival prediction, yet modality incompleteness is pervasive in real-world medical data and undermines the stability and applicability of existing models. To address this challenge, we propose DMG-PoE, a robust learning framework for multi-omics survival prediction. The framework introduces a dynamic modality graph (DMG) that propagates complementary information across modality nodes to compensate for missing inputs, combined with an uncertainty-aware product-of-experts (PoE) fusion mechanism that adaptively weights each modality by its estimated precision, and a consistency-constrained training strategy that stabilizes predictions across varying modality subsets. Systematic evaluation on the SurvBoard benchmark across 10 TCGA cancer types and seven omics modalities demonstrates that DMG-PoE achieves the highest average C-index under both the standard setting (complete-modality training/testing) and the missing-augmented setting (training additionally includes real incomplete samples and nonclinical modality dropout, with complete-only testing); controlled per-modality masking experiments validate effective test-time recovery in 22/24 cancer-modality combinations, and cross-cohort transfer with only two of seven modalities available further confirms generalization under severe input restriction.

Related

  1. TransSurv: compatibility-aware transfer learning for target-cohort survival prediction
    Zheng Qu, Hongzhe Zhang, Katherine L Nathanson +3
    Briefings in BioinformaticsPeer-reviewedCC BY 4.0Full textMolecular biology & gene regulationPMC
  2. ESMDynamic: Fast and accurate prediction of protein dynamic contact maps from single sequences
    Diego E. Kleiman, Jiangyan Feng, Zhengyuan Xue +1
    Nature CommunicationsPeer-reviewedCC BY-NC-ND7 citedMolecular biology & gene regulationOpenAlex
  3. Dynamic graph prior-guided latent adaptation for generalizable single-cell perturbation response prediction.
    Yunlong Bie, Peishan Fan, Qijin Yin +1
    Bioinformatics (Oxford, England)Peer-reviewedCC BYFull textMolecular biology & gene regulationEurope PMCCrossref
  4. Associations Between Physical Activity Measurement Modality and Gut Microbiome in Breast Cancer Survivors
    Whitney N. Neal, Keri M. Kemp, Rebecca B. Little +11
    NutrientsPeer-reviewedCC BYFull textMolecular biology & gene regulationEurope PMCCrossref
  5. Metabolic reprogramming and metabolic homeostasis in podocyte injury: novel mechanistic insights and natural product therapies
    Zhongjie Qu, Wenqing Ba, Meijing Tang +3
    Frontiers in cell and developmental biologyPeer-reviewedCC BYFull textGeneral biochemistryEurope PMC

Cite

BibTeXRIS
BibTeX
@article{xie2026multiomics,
  title = {{Multi-omics survival prediction under modality incompleteness via dynamic modality graphs and product-of-experts fusion.}},
  author = {Weidong Xie and Huixia Zhang and Jinzhu Yang and Kun Yu and Wei Li},
  journal = {Briefings in bioinformatics},
  year = {2026},
  month = nov,
  doi = {10.1093/bib/bbag596},
  url = {https://europepmc.org/article/MED/42969853},
  note = {CC BY-NC},
}
RIS
TY  - JOUR
TI  - Multi-omics survival prediction under modality incompleteness via dynamic modality graphs and product-of-experts fusion.
AU  - Weidong Xie
AU  - Huixia Zhang
AU  - Jinzhu Yang
AU  - Kun Yu
AU  - Wei Li
PY  - 2026
DA  - 2026/11/01
JO  - Briefings in bioinformatics
DO  - 10.1093/bib/bbag596
UR  - https://europepmc.org/article/MED/42969853
AB  - Multi-omics data integration provides rich complementary information for cancer survival prediction, yet modality incompleteness is pervasive in real-world medical data and undermines the stability and applicability of existing models. To address this challenge, we propose DMG-PoE, a robust learning framework for multi-omics survival prediction. The framework introduces a dynamic modality graph (DMG) that propagates complementary information across modality nodes to compensate for missing inputs, combined with an uncertainty-aware product-of-experts (PoE) fusion mechanism that adaptively weights each modality by its estimated precision, and a consistency-constrained training strategy that stabilizes predictions across varying modality subsets. Systematic evaluation on the SurvBoard benchmark across 10 TCGA cancer types and seven omics modalities demonstrates that DMG-PoE achieves the highest average C-index under both the standard setting (complete-modality training/testing) and the missing-augmented setting (training additionally includes real incomplete samples and nonclinical modality dropout, with complete-only testing); controlled per-modality masking experiments validate effective test-time recovery in 22/24 cancer-modality combinations, and cross-cohort transfer with only two of seven modalities available further confirms generalization under severe input restriction.
SN  - 1467-5463
SN  - 1477-4054
ER  -