![[2602.14177] Towards Spatial Transcriptomics-driven Pathology Foundation Models](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.14177] Towards Spatial Transcriptomics-driven Pathology Foundation Models
Summary
This article presents Spatial Expression-Aligned Learning (SEAL), a framework that integrates spatial transcriptomics with pathology models to enhance gene expression analysis and improve predictive performance in medical imaging.
Why It Matters
The integration of spatial transcriptomics with pathology foundation models represents a significant advancement in medical diagnostics. By improving the accuracy of gene expression predictions, this research could lead to better patient outcomes and more personalized treatment strategies, making it highly relevant in the fields of pathology and computational biology.
Key Takeaways
- SEAL enhances existing pathology models by incorporating localized molecular data.
- The framework shows improved performance in predicting molecular status and treatment responses.
- Robust domain generalization allows SEAL to perform well on out-of-distribution evaluations.
- Cross-modal capabilities enable innovative applications like gene-to-image retrieval.
- The approach offers a practical method for augmenting pathology models without extensive retraining.
Computer Science > Computer Vision and Pattern Recognition arXiv:2602.14177 (cs) [Submitted on 15 Feb 2026] Title:Towards Spatial Transcriptomics-driven Pathology Foundation Models Authors:Konstantin Hemker, Andrew H. Song, Cristina Almagro-Pérez, Guillaume Jaume, Sophia J. Wagner, Anurag Vaidya, Nikola Simidjievski, Mateja Jamnik, Faisal Mahmood View a PDF of the paper titled Towards Spatial Transcriptomics-driven Pathology Foundation Models, by Konstantin Hemker and 8 other authors View PDF HTML (experimental) Abstract:Spatial transcriptomics (ST) provides spatially resolved measurements of gene expression, enabling characterization of the molecular landscape of human tissue beyond histological assessment as well as localized readouts that can be aligned with morphology. Concurrently, the success of multimodal foundation models that integrate vision with complementary modalities suggests that morphomolecular coupling between local expression and morphology can be systematically used to improve histological representations themselves. We introduce Spatial Expression-Aligned Learning (SEAL), a vision-omics self-supervised learning framework that infuses localized molecular information into pathology vision encoders. Rather than training new encoders from scratch, SEAL is designed as a parameter-efficient vision-omics finetuning method that can be flexibly applied to widely used pathology foundation models. We instantiate SEAL by training on over 700,000 paired gene express...
