PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
March 3, 20260 citations

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

XZXianghao ZhanJXJingyu XuYZYuanning Zheng

Key Points

  • SAGE-FM achieves 91% correlation for masked gene recovery, showcasing high efficacy in spatial transcriptomics.
  • Trained on 416 human Visium samples across 15 organs, the model outperforms existing methods in unsupervised clustering.
  • Analysis utilizes graph convolutional networks for creating spatially coherent embeddings in transcriptomics.
  • Significantly enhances pathologist-defined annotations and glioblastoma subtype predictions, indicating its potential applications.

Abstract

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 (GCN) trained with a masked-central-spot prediction objective. Trained on 416 human Visium samples spanning 15 organs, SAGE-FM learns spatially coherent embeddings that recover masked genes robustly, with 91% of masked genes showing significant correlations (p In silico perturbation experiments further show 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.

Ask AI
Helpful
Bookmark
Share

Cite This Study

Zhan et al. (2026) studied this question.

synapsesocial.com/papers/69a7684abadf0bb9e87e4411
Ask AI
Helpful
Bookmark
Share