Decoding spatial transcriptomics across multicellular and subcellular resolutions

2026-05-18 · Source: Machine learning : nature.com subject feeds · Field: Science & Research — Life Sciences & Biology, Health & Medical Research, Artificial Intelligence & Machine Learning · Depth: Expert, medium

Summary

STARS (Spatial Transcriptomics across Resolutions for Single Cells) is a new computational method designed to reconstruct single-cell-level gene expression from spatial transcriptomics (ST) platforms, which typically operate at multicellular or subcellular spot levels. Current ST platforms like Visium, Visium HD, and Stereo-seq struggle with accurately decomposing spots into single cells, the fundamental biological units. STARS addresses this by combining high-resolution histology images with spot-level transcriptomics data, leveraging a Vision Transformer model and contrastive learning. The method was validated using in-house mouse lung datasets from Visium, Visium HD, and Stereo-seq, as well as public datasets. STARS effectively identifies regions of interest at the tissue level, distinguishes immune cell subtypes (e.g., CD4/CD8 T cells, CAF subtypes, *SPP1*+ macrophages) at the individual cell level, and improves cell type separation and differential gene expression analysis at the molecular level.

Key takeaway

For AI Scientists and Research Scientists working with spatial transcriptomics, STARS offers a crucial advancement by enabling single-cell gene expression reconstruction. This capability allows for more precise identification of cell types and their spatial organization, which is vital for understanding complex biological processes and disease mechanisms. You should consider integrating STARS into your analysis pipeline to achieve higher resolution insights from existing and future ST datasets.

Key insights

STARS reconstructs single-cell gene expression from spatial transcriptomics data using Vision Transformers and contrastive learning.

Principles

• Combine histology with transcriptomics for resolution enhancement.
• Single-cell resolution improves biological insight.

Method

STARS utilizes a Vision Transformer model and contrastive learning to integrate high-resolution histology images with spot-level transcriptomics data, enabling reconstruction of single-cell gene expression.

In practice

• Identify specific tissue structures and immune regions.
• Distinguish rare immune cell subtypes like *SPP1*+ macrophages.

Topics

• Spatial Transcriptomics
• Single-cell Gene Expression
• STARS Method
• Vision Transformer
• Contrastive Learning

Best for: AI Scientist, Research Scientist

Related on AIssential

• Identifying spatial single-cell-level interactions with graph transformer — A self-supervised graph transformer identifies spatial single-cell interactions without known ligand-receptor pairs.
• CellNiche represents cellular microenvironments in atlas-scale spatial omics data with contrastive learning — CellNiche uses contrastive learning on spatial omics data to map cellular microenvironments across diverse biological contexts.
• Whole mouse mapping — New spatial genomics tools enable whole-body transcriptomics and high-resolution cell type identification in mice.
• Latent space mapping of interpretable structural coordinates from stochastic single-molecule signals — Latent-space mapping resolves stochastic signal warping in nanopore sensors, enabling interpretable molecular coordinate systems.
• scGTN: Deep Siamese Graph Transformer Network for Single-cell RNA Sequencing Clustering — scGTN integrates gene expression and intercellular structure via a Siamese graph transformer for robust scRNA-seq clustering.
• STAR: SpatioTemporal Adaptive Reward Allocation for Text-to-Image RL Post-Training — STAR adaptively allocates rewards spatio-temporally in text-to-image RL post-training, focusing policy updates on critical image regions.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine learning : nature.com subject feeds.