Reconstructing developmental and disease progression with sample-level embeddings.

scSLIDE embeds single-cell samples into density-profile vectors that reveal separable axes of COVID-19 infection and severity, plus disease trajectories and embryonic pseudostages.

• Why it matters: Case-control single-cell analyses can collapse meaningful within-group sample variation into binary labels, limiting comparisons needed for progression and coordinated programs.
• What they did: scSLIDE (single-cell sample-level inference using density estimation) converts each sample’s cells into a compact high-dimensional density profile, then compares profiles to compute sample embeddings; implemented in Seurat v6.
• The result: Applied to COVID-19, Alzheimer’s disease, and zebrafish embryogenesis, the embeddings identify patient subgroups, reconstruct disease trajectories, and uncover coordinated cellular programs across samples.

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