Spatially guided translation from histology images to transcriptomic profiles using foundation model-driven contrastive learning
- 1Department of Biochemistry, Western University, London, Canada
- 2Department of Computer Sciences, Western University, London, Canada
- 3Division of Biostatistics, University of Toronto, Toronto, Canada
Abstract
Spatial transcriptomics (ST) enhances single-cell RNA sequencing by revealing transcript distribution, offering critical insights into heterogeneous diseases such as breast cancer. However, the high cost and lengthy processes of generating high-quality ST data limit clinical application. Recent deep learning methods predict ST from histology images, but often fail to capture both morphological features and spatial context. We introduce FOCST, a foundation model-driven framework for ST imputation that leverages spatial guided contrastive learning. FOCST begins with UNI, a large histopathology foundation model, to extract visual features from tissue images. These are integrated with expression data in a unified embedding space via contrastive learning, enabling cross-modal prediction and imputation. To further enhance spatial awareness, a graph neural network incorporates positional information, improving regional detection and interpretability.Benchmarking demonstrates FOCST’s superior performance over state-of-the-art methods and alternative vision encoders (paired Wilcoxon signed-rank tests, FDR-adjusted p < 0.05, N = 6 images). Predicted profiles enable clinically relevant downstream analyses, including patient stratification by treatment response (ROC AUC (Receiver Operating Characteristic – Area Under the Curve) = 0.79). Our results highlight the promise of combining foundation models and spatially guided learning to efficiently generate ST insights, advancing cancer research and precision medicine. Author summary: Spatial transcriptomics offers powerful insights into tissue heterogeneity but remains expensive and impractical for large-scale clinical use. We developed FOCST, a foundation model-based approach that predicts transcriptomic profiles from standard histology images using spatially guided contrastive learning. By leveraging UNI, a large histopathology model, and a graph neural network to encode spatial context, FOCST produces highly accurate and interpretable ST predictions. Compared to existing methods, it achieves superior accuracy and supports downstream analyses such as patient treatment response prediction. FOCST demonstrates the potential of foundation models to bring cost-effective, scalable spatial biology into cancer research and clinical translation.
Related
- Slide-level batch structure limits histology-guided supervision of transcriptomic foundation models
- A multimodal foundation model to integrate histology with spatial molecular programs
- Concept-aware contrastive representation learning for cross-translation semantic consistency modeling of the Analects
- Data distribution impacts the performance and generalisability of contrastive learning-based foundation models of electrocardiograms
- SVC: A Vision Transformer-based Spatially and Virtually Embedded Cell Model for Deciphering Subcellular Spatial Transcriptomic Heterogeneity
BibTeX
@article{huang2026spatially,
title = {{Spatially guided translation from histology images to transcriptomic profiles using foundation model-driven contrastive learning}},
author = {Zi Huai Huang and Ziyang Xu and Pingzhao Hu},
journal = {PLOS Computational Biology},
year = {2026},
month = sep,
doi = {10.1371/journal.pcbi.1014762},
url = {https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1014762},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Spatially guided translation from histology images to transcriptomic profiles using foundation model-driven contrastive learning
AU - Zi Huai Huang
AU - Ziyang Xu
AU - Pingzhao Hu
PY - 2026
DA - 2026/09/08
JO - PLOS Computational Biology
DO - 10.1371/journal.pcbi.1014762
UR - https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1014762
AB - Spatial transcriptomics (ST) enhances single-cell RNA sequencing by revealing transcript distribution, offering critical insights into heterogeneous diseases such as breast cancer. However, the high cost and lengthy processes of generating high-quality ST data limit clinical application. Recent deep learning methods predict ST from histology images, but often fail to capture both morphological features and spatial context. We introduce FOCST, a foundation model-driven framework for ST imputation that leverages spatial guided contrastive learning. FOCST begins with UNI, a large histopathology foundation model, to extract visual features from tissue images. These are integrated with expression data in a unified embedding space via contrastive learning, enabling cross-modal prediction and imputation. To further enhance spatial awareness, a graph neural network incorporates positional information, improving regional detection and interpretability.Benchmarking demonstrates FOCST’s superior performance over state-of-the-art methods and alternative vision encoders (paired Wilcoxon signed-rank tests, FDR-adjusted p < 0.05, N = 6 images). Predicted profiles enable clinically relevant downstream analyses, including patient stratification by treatment response (ROC AUC (Receiver Operating Characteristic – Area Under the Curve) = 0.79). Our results highlight the promise of combining foundation models and spatially guided learning to efficiently generate ST insights, advancing cancer research and precision medicine. Author summary: Spatial transcriptomics offers powerful insights into tissue heterogeneity but remains expensive and impractical for large-scale clinical use. We developed FOCST, a foundation model-based approach that predicts transcriptomic profiles from standard histology images using spatially guided contrastive learning. By leveraging UNI, a large histopathology model, and a graph neural network to encode spatial context, FOCST produces highly accurate and interpretable ST predictions. Compared to existing methods, it achieves superior accuracy and supports downstream analyses such as patient treatment response prediction. FOCST demonstrates the potential of foundation models to bring cost-effective, scalable spatial biology into cancer research and clinical translation.
SN - 1553-7358
ER -