SAGE-FM: A lightweight and interpretable spatial transcriptomics foundation model

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhan, Xianghao, Xu, Jingyu, Zheng, Yuanning, Good, Zinaida, Gevaert, Olivier
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917216804405248
author Zhan, Xianghao
Xu, Jingyu
Zheng, Yuanning
Good, Zinaida
Gevaert, Olivier
author_facet Zhan, Xianghao
Xu, Jingyu
Zheng, Yuanning
Good, Zinaida
Gevaert, Olivier
contents Spatial transcriptomics enables spatial gene expression profiling, motivating computational models that capture spatially conditioned regulatory relationships. We introduce SAGE-FM, a lightweight spatial transcriptomics foundation model based on graph convolutional networks (GCNs) trained with a masked central spot prediction objective. Trained on 416 human Visium samples spanning 15 organs, SAGE-FM learns spatially coherent embeddings that robustly recover masked genes, with 91% of masked genes showing significant correlations (p < 0.05). The embeddings generated by SAGE-FM outperform MOFA and existing spatial transcriptomics methods in unsupervised clustering and preservation of biological heterogeneity. SAGE-FM generalizes to downstream tasks, enabling 81% accuracy in pathologist-defined spot annotation in oropharyngeal squamous cell carcinoma and improving glioblastoma subtype prediction relative to MOFA. In silico perturbation experiments further demonstrate that the model captures directional ligand-receptor and upstream-downstream regulatory effects consistent with ground truth. These results demonstrate that simple, parameter-efficient GCNs can serve as biologically interpretable and spatially aware foundation models for large-scale spatial transcriptomics.
format Preprint
id arxiv_https___arxiv_org_abs_2601_15504
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle SAGE-FM: A lightweight and interpretable spatial transcriptomics foundation model
Zhan, Xianghao
Xu, Jingyu
Zheng, Yuanning
Good, Zinaida
Gevaert, Olivier
Machine Learning
Genomics
Quantitative Methods
Spatial transcriptomics enables spatial gene expression profiling, motivating computational models that capture spatially conditioned regulatory relationships. We introduce SAGE-FM, a lightweight spatial transcriptomics foundation model based on graph convolutional networks (GCNs) trained with a masked central spot prediction objective. Trained on 416 human Visium samples spanning 15 organs, SAGE-FM learns spatially coherent embeddings that robustly recover masked genes, with 91% of masked genes showing significant correlations (p < 0.05). The embeddings generated by SAGE-FM outperform MOFA and existing spatial transcriptomics methods in unsupervised clustering and preservation of biological heterogeneity. SAGE-FM generalizes to downstream tasks, enabling 81% accuracy in pathologist-defined spot annotation in oropharyngeal squamous cell carcinoma and improving glioblastoma subtype prediction relative to MOFA. In silico perturbation experiments further demonstrate that the model captures directional ligand-receptor and upstream-downstream regulatory effects consistent with ground truth. These results demonstrate that simple, parameter-efficient GCNs can serve as biologically interpretable and spatially aware foundation models for large-scale spatial transcriptomics.
title SAGE-FM: A lightweight and interpretable spatial transcriptomics foundation model
topic Machine Learning
Genomics
Quantitative Methods
url https://arxiv.org/abs/2601.15504