Related Experiment Video
Updated: Jun 10, 2025
03:37
Author Spotlight: Impact of Intergenic Interactions on Disease-Identifying Dark Biomarkers
Published on: March 1, 2024
647
Gene expression prediction from histology images via hypergraph neural networks
Bo Li1, Yong Zhang1, Qing Wang2
1Beijing Key Laboratory of Multimedia and Intelligent Software Technology, Faculty of Information Technology, Beijing Institute of Artificial Intelligence, Beijing University of Technology, Beijing, 100124, China.
Briefings in Bioinformatics
|October 14, 2024
Summary
This study introduces HGGEP, a new hypergraph neural network model for predicting gene expression from histology images. HGGEP improves accuracy by considering cell morphology and multi-stage image features, outperforming existing methods in cancer and tumor datasets.
Area of Science:
- Computational biology
- Genomics
- Biomedical imaging
Background:
- Spatial transcriptomics provides gene distribution insights for biology and disease research.
- Predicting gene expression from histology images is cost-effective but challenging.
- Current methods overlook cell morphology-gene expression links and multi-stage image features.
Purpose of the Study:
- To develop a novel model, HGGEP, for accurate gene expression prediction from histology images.
- To address limitations of existing methods by integrating cell morphology and multi-stage features.
Main Methods:
- Proposed a hypergraph neural network (HGGEP) model.
- Incorporated a gradient enhancement module for cell morphology perception.
- Utilized a lightweight backbone and attention mechanisms for multi-stage feature extraction and refinement.
- Employed hypergraph construction to capture higher-order associations among features at different scales.
Main Results:
- HGGEP demonstrated superior performance compared to existing methods.
- The model effectively predicted gene expression from histology images.
- Experiments were conducted on diverse datasets, including cancer and tumor samples.
Conclusions:
- HGGEP offers a powerful approach for gene expression prediction from histology images.
- The model's ability to leverage cell morphology and multi-stage features enhances predictive accuracy.
- This method holds promise for advancing biological and disease research through cost-effective gene expression analysis.
