CloneVI: A deep generative model for deciphering the relationships between cell state and mitochondrial genotype

Data and fitted models of the article "CloneVI: A deep generative model for deciphering the relationships between cell state and mitochondrial genotype" (Hibi et al.). The code that reproduces the figures and numbers of the article from these files is included in the repository at https://github.com/TaichiHIBI/cloneVI. Files cloneVI_data.tar.gz: the simulated data of Fig. 2; the model inputs (h5ad: cell-state counts and mtDNA allele counts) of the MAESTER, CLL mtscATAC-seq and ovarian tumor mtscATAC-seq datasets; the cell-type labels of the source study matched to the cells of the ovarian cohort; the chromVAR motif activity of every cell of the ovarian cohort. cloneVI_output.tar.gz: the five CloneVI fits analysed in the article (results as h5ad, model weights, settings, train/validation/test split). cohort_Ovarian_atac.h5seurat: accessibility of the ovarian cohort over all peaks (Signac object), needed only to recompute the motif activity. The fragment files are not included. SHA256SUMS.txt: SHA-256 of every file after unpacking. Usage. At the root of a clone of the repository:tar -xzf cloneVI_data.tar.gztar -xzf cloneVI_output.tar.gzmv cohort_Ovarian_atac.h5seurat data/mitotrek/sha256sum -c SHA256SUMS.txt The content of each file and the two files to download from their sources (the PBMC3k counts of 10x Genomics and the CLL fragments of https://osf.io/bupge/) are described in data/README.md and output/README.md of the repository. Sources. The model inputs are derived from public data: the MAESTER data of Miller et al. (2022, Nat Biotechnol 40, 1030–1034; GEO GSE182685), the CLL mtscATAC-seq data of Lareau et al. (2021, Nat Biotechnol 39, 451–461; https://osf.io/bupge/), and the ovarian tumor mtscATAC-seq data of Liu et al. (2026, Cancer Cell 44, 1509–1521.e4; GEO GSE302113), reprocessed from the raw libraries. The simulated data are drawn from the PBMC3k dataset of 10x Genomics. Please cite the source studies together with the article when using the derived data.