NEOGRAN: traceable graph-text fusion for disease–protein relation prediction in biomedical knowledge graphs
- 1College of Computer Science and Artificial Intelligence, Northeast Forestry University, Harbin, 150040 China
- 2Department of Computer Science and Technology, Harbin Institute of Technology, Harbin, 150001 China
Abstract
Background: Accurately predicting disease–protein relations in biomedical knowledge graphs helps link disease phenotypes to molecular mechanisms and supports disease-related knowledge discovery and candidate target identification. Biomedical knowledge graphs organize multisource biomedical knowledge, including diseases, proteins, drugs, and pathways, as entity nodes and relational edges, providing a structured foundation for modeling complex biomedical associations. Existing methods are often constrained by single-modality modeling, shallow graph–text fusion, and insufficient traceable evidence, which limits their ability to exploit graph–text complementarity and weakens downstream validation and structural evidence interpretation.
Results: To address these limitations, we propose NEOGRAN, a graph–text collaborative framework comprising three core modules for relation prediction in biomedical knowledge graphs. The dual-encoder architecture captures graph structural patterns and biomedical entity representations to mitigate single-modality modeling. The bidirectional cross-attention module enables deep graph–text interaction to overcome shallow fusion. The interpretable path module generates traceable evidence paths to support prediction verification, structural evidence interpretation, and hypothesis generation. On PrimeKG, NEOGRAN achieved an AUPR of 0.9860 and an AUROC of 0.9875 under the 1:1 sampled classification setting, and further obtained an MRR of 0.0735 in the all-candidate ranking evaluation. External validation on BioKG further shows that NEOGRAN remains effective under differences in entity coverage, relation composition, and local topology, supporting its method-level generalizability across knowledge graph sources.
Conclusions: NEOGRAN provides an effective solution for relation prediction in biomedical knowledge graphs while offering traceable structural evidence for hypothesis generation and further biological validation.
Related
- Relation-aware multimodal knowledge graph learning for drug–drug interaction prediction
- UKB-KG: Knowledge Graph for Integrating and Enhancing Biomedical Insights from the UK Biobank
- Prospective In-silico Simulation of the VESALIUS-CV Trial Using Biomedical Knowledge Graph and Real-World Data-Driven AI Modeling
- Recent Advances in Leveraging AI for Scalable and Extensible Biomedical Knowledge Management
- Knowledge, Awareness, and Perception toward Ethical Guidelines for Artificial Intelligence in Biomedical Research and Healthcare among Dental Faculty and Postgraduates – A Cross-sectional Study
BibTeX
@article{wang2026neogran,
title = {{NEOGRAN: traceable graph-text fusion for disease–protein relation prediction in biomedical knowledge graphs}},
author = {Zhenxing Wang and Qihe Wang and Murong Zhou and Guohua Wang and Yuming Zhao},
journal = {BMC Bioinformatics},
year = {2026},
month = aug,
doi = {10.1186/s12859-026-06580-3},
url = {https://doi.org/10.1186/s12859-026-06580-3},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - NEOGRAN: traceable graph-text fusion for disease–protein relation prediction in biomedical knowledge graphs
AU - Zhenxing Wang
AU - Qihe Wang
AU - Murong Zhou
AU - Guohua Wang
AU - Yuming Zhao
PY - 2026
DA - 2026/08/08
JO - BMC Bioinformatics
DO - 10.1186/s12859-026-06580-3
UR - https://doi.org/10.1186/s12859-026-06580-3
AB - Background: Accurately predicting disease–protein relations in biomedical knowledge graphs helps link disease phenotypes to molecular mechanisms and supports disease-related knowledge discovery and candidate target identification. Biomedical knowledge graphs organize multisource biomedical knowledge, including diseases, proteins, drugs, and pathways, as entity nodes and relational edges, providing a structured foundation for modeling complex biomedical associations. Existing methods are often constrained by single-modality modeling, shallow graph–text fusion, and insufficient traceable evidence, which limits their ability to exploit graph–text complementarity and weakens downstream validation and structural evidence interpretation. Results: To address these limitations, we propose NEOGRAN, a graph–text collaborative framework comprising three core modules for relation prediction in biomedical knowledge graphs. The dual-encoder architecture captures graph structural patterns and biomedical entity representations to mitigate single-modality modeling. The bidirectional cross-attention module enables deep graph–text interaction to overcome shallow fusion. The interpretable path module generates traceable evidence paths to support prediction verification, structural evidence interpretation, and hypothesis generation. On PrimeKG, NEOGRAN achieved an AUPR of 0.9860 and an AUROC of 0.9875 under the 1:1 sampled classification setting, and further obtained an MRR of 0.0735 in the all-candidate ranking evaluation. External validation on BioKG further shows that NEOGRAN remains effective under differences in entity coverage, relation composition, and local topology, supporting its method-level generalizability across knowledge graph sources. Conclusions: NEOGRAN provides an effective solution for relation prediction in biomedical knowledge graphs while offering traceable structural evidence for hypothesis generation and further biological validation.
SN - 1471-2105
ER -