Preprocessed Mammalian datasets

Preprocessed single-cell RNA-seq data from mammalian and Drosophila circadian studies This deposit contains preprocessed single-cell/single-nucleus RNA-seq data from four previously published circadian studies, curated into AnnData (.h5ad) objects for circadian phase-inference analysis. It comprises two files: 1. mammalian_datasets.h5ad — three mammalian tissues, each labeled in obs["study"]: Liver (Droin_2021) — Droin, C. et al., Nature Metabolism 3, 43–58 (2021). https://doi.org/10.1038/s42255-020-00323-1 Aorta (Auerbach_2022) — Auerbach, B. J., FitzGerald, G. A. & Li, M., Nature Communications 13, 6580 (2022). https://doi.org/10.1038/s41467-022-34185-w Skin (Duan_2024) — Duan, J. et al. "tauFisher predicts circadian time from a single sample of bulk and single-cell transcriptomic data." Nature Communications 15, 3840 (2024). https://doi.org/10.1038/s41467-024-48041-6 2. drosophila_dataset.h5ad — Drosophila clock neurons (obs["study"] = Ma_2021): Ma, D. et al. "A transcriptomic taxonomy of Drosophila circadian neurons around the clock." eLife 10, e63056 (2021). https://doi.org/10.7554/eLife.63056 Structure. In both files each cell carries its source study in obs["study"], cell-type annotations in obs["celltype"], and time-of-day metadata (e.g. ZT). The Drosophila file additionally records the light regime in obs["light_condition"] (CT/ZT). Count layers. Raw UMI counts are stored in two layers, spliced and unspliced. Only the spliced layer was used in our analyses; the .X matrix should be ignored, and unspliced is provided for completeness. Analysis code: https://github.com/AndreaSalati/scRITMO — Associated preprint: https://doi.org/10.64898/2026.03.30.715278