Comparative Biology at Single-Cell Resolution: Rigorous Matching of Atlases for Cross-Species Analysis

Single-cell transcriptomics has revolutionised developmental biology by providing an unprecedented, fine-grained view of cellular lineages. However, our ability to compare species and distinguish universal from species-specific developmental principles remains limited by biological and technical variability. To address this, we introduce RIMA (RIgorous Matching of Atlases), a method for quantitatively comparing transcriptomic atlases across species at near-single-cell resolution. RIMA uses a novel computational approach to identify matching cell states across atlases and leverages this to enable quantitative comparative analyses. Applied to gastrulation in mouse, rabbit, and macaque, RIMA recapitulates a developmental hourglass pattern, identifying a molecular similarity bottleneck at the onset of organogenesis. It further uncovers conserved developmental programmes, including a core set of transcription factors driving epithelial-mesenchymal transition, and highlights transcriptional boosts of erythroid differentiation genes that are conserved across species but exhibit shifted onset timing. Furthermore, RIMA enables cross-species prediction of gene expression, augmenting sparse atlases and correcting differences between model and target organisms. Beyond cross-species comparisons, RIMA’s framework extends naturally to any setting where large systematic differences exist across datasets, including in vitro to in vivo comparisons, opening new avenues for improving biological models and advancing translational research.