TRACER: Twin-contrastive learning for fine-grained and rare spatial domain identification in spatial transcriptomics.

TRACER, a twin-contrastive learning framework for spatial transcriptomics, improved fine-grained and rare spatial domain identification across multiple datasets, including human dorsolateral prefrontal cortex (10x Visium).

• Why it matters: Fine-grained and rare spatial domains are hard to detect because they have few spots, subtle transcriptional differences, and strong dependence on neighboring regions.
• What they did: TRACER combines a graph convolutional autoencoder with neighborhood-aware instance-level and cluster-level contrastive learning to preserve local spatial continuity while increasing domain separability.
• The result: Across human and mouse spatial transcriptomic datasets (including STARmap, osmFISH, and BARISTAseq), TRACER recovered major cortical/anatomical structures, improved rare or transitional domains in mouse somatosensory cortex, and found candidate small functionally distinct domains within a broadly annotated human breast cancer region.

AI summary of the abstract · may contain errors · not medical advice · read the paper.