Related Experiment Video
Updated: Sep 15, 2025
09:19
Spatial Profiling of Protein and RNA Expression in Tissue: An Approach to Fine-Tune Virtual Microdissection
Published on: July 6, 2022
5.0K
DGAT: A Dual-Graph Attention Network for Inferring Spatial Protein Landscapes from Transcriptomics
Haoyu Wang1, Brittany Cody2, Hatice Ulku Osmanbeyoglu1
1Department of Biomedical Informatics, University of Pittsburgh School of Medicine, Pittsburgh, PA, USA.
Biorxiv : the Preprint Server for Biology
|July 17, 2025
Summary
We developed DGAT, a deep learning tool that predicts protein levels from spatial transcriptomics data. This method enhances understanding of cellular function and tissue organization by revealing protein expression patterns previously hidden in transcriptomics-only studies.
Area of Science:
- Spatial omics
- Computational biology
- Biotechnology
Background:
- Spatial transcriptomics (ST) offers genome-wide mRNA profiles within tissue context.
- ST lacks direct protein-level measurements crucial for cellular function and microenvironment interpretation.
Purpose of the Study:
- To develop a deep learning framework, DGAT (Dual-Graph Attention Network), for imputing spatial protein expression from transcriptomics-only ST data.
- To leverage RNA-protein relationships learned from spatial CITE-seq datasets for accurate protein imputation.
Main Methods:
- DGAT constructs heterogeneous graphs integrating transcriptomic, proteomic, and spatial information.
- Graph attention networks encode integrated data, with task-specific decoders reconstructing mRNA and predicting protein abundance from a shared latent representation.
Main Results:
- DGAT demonstrates superior protein imputation accuracy compared to existing methods across diverse datasets (tonsil, breast cancer, glioblastoma, malignant mesothelioma).
- Application to ST datasets reveals spatially distinct cell states, immune phenotypes, and tissue architectures not apparent from transcriptomics alone.
Conclusions:
- DGAT enables proteome-level insights from transcriptomics-only spatial omics data.
- This approach bridges a critical gap in spatial omics, enhancing functional interpretation in cancer, immunology, and precision medicine.
