LEAF-1 pretraining corpus: scATAC-seq fragments

This record contains the single-cell ATAC-seq (scATAC-seq) portion of the pretraining corpus of LEAF-1. Paper: https://www.biorxiv.org/content/10.64898/2026.07.09.737627v1 Other portions of the LEAF-1 pretraining corpus:cfDNA fragments: huggingface.co/datasets/jing-z/leaf1-pretraining-cfdna (sample manifest and checksums: 10.5281/zenodo.23195009)TCGA ATAC-seq (controlled access; sample manifest and processing code): 10.5281/zenodo.23195360 91 pseudo-bulk samples in 92 bgzip-compressed BED files (one sample, UMB4540 frontal cortex, has two replicate files). Fragments derive from the CATlas atlas of Zhang et al., 2021, Cell (https://doi.org/10.1016/j.cell.2021.10.024), filtered for length ≥ 20 bp, hg38 reference chromosomes, and the ENCODE hg38 blacklist v2. Please cite Zhang et al. (2021) when reusing these files. See README.md in this record for file formats, the index sidecars, verification checksums and full provenance.