The abstract is read at the publisher; the summary is JClub's.
CIRCA learns identifiable cell- and neighborhood-level transcriptional programs and directed interactions from spatial transcriptomics, achieving highest biological accuracy in six of seven cell-resolution datasets.
- Why it matters: Spatial transcriptomics can map cell–microenvironment relationships in situ, but conventional nonlinear models are non-identifiable, limiting interpretable program recovery and interaction inference.
- What they did: CIRCA is a self-supervised causal representation learning framework that jointly infers identifiable cell and niche programs plus directed interactions under explicit assumptions, benchmarked against five methods across seven datasets.
- The result: In an AD mouse model, adding amyloid-{beta} pathology identifies four region-specific plaque niches enriched for microglia, with region-varying disease-response strength and hippocampal/cortical enrichment of AD genetic risk.
The findingWhy it mattersWhat they didThe result
AI summary of the abstract · may contain errors · not medical advice · read the paper.