Related Experiment Video
Updated: Aug 6, 2026
10:16
Mining Spatial Transcriptomics Datasets using DeepSpaceDB
Published on: September 5, 2025
Deciphering spatial heterogeneity by multimodal spatial transcriptomics modelling with SpatialModal
Xingyi Li1,2, Dongmin Zhao1, Xiangting Jia1
1School of Computer Science, Northwestern Polytechnical University, Shaanxi, China.
Bioinformatics (Oxford, England)
|July 21, 2026
Summary
SpatialModal, a new multimodal graph learning framework, effectively integrates spatial transcriptomics data. It reveals complex biological patterns across various tissues and diseases, even with limited data.
Area of Science:
- Genomics
- Computational Biology
- Bioinformatics
Background:
- Spatial transcriptomics (ST) enables joint analysis of gene expression and histology with spatial coordinates.
- ST offers opportunities to study tissue heterogeneity but lacks effective multimodal data integration methods.
Purpose of the Study:
- To develop a computational framework for synergistic modeling of multimodal spatial transcriptomics data.
- To enhance the understanding of spatial heterogeneity in complex biological tissues.
Main Methods:
- Proposed SpatialModal, a multimodal graph learning framework.
- Employed a hierarchical representation strategy and dual-level contrastive learning.
- Validated across diverse human and mouse ST datasets.
Main Results:
- SpatialModal effectively integrates multimodal ST data for robust joint representation learning.
- Demonstrated capability in analyzing brain architecture, tumor microenvironments, Alzheimer's disease, and embryonic development.
- Showcased versatility and robustness, performing well on unimodal datasets.
Conclusions:
- SpatialModal provides a powerful tool for deciphering spatial heterogeneity in biological tissues.
- The framework is broadly applicable across various ST platforms and research areas.
- Enables deeper insights into complex biological systems through integrated multimodal analysis.
