GM12878_dnase-data

Dataset Summary

This dataset contains genomic intervals used to train a regression model on GM12878 DNase data, described in Lal et al. 2025 (https://www.nature.com/articles/s41592-025-02868-z). Genome coordinates correspond to the hg38 reference genome.

Repository Content

The repository includes one BED file and one Jupyter notebook:

1. intervals.bed: Genomic intervals stored in BED format.
2. 1_process_GM12878_data.ipynb: Jupyter notebook containing the preprocessing steps used to generate the intervals.bed file.

Dataset Structure

Statistics

• Number of intervals: 435,055
• Interval length: 2,114 bp (all intervals)
• Genome build: hg38

Intervals file (intervals.bed)

BED format (tab-separated). There are three columns with no header:

• Chromosome name
• Start position
• End position

Usage

from huggingface_hub import hf_hub_download
import pandas as pd

file_path = hf_hub_download(
    repo_id="Genentech/GM12878_dnase-data",
    filename="intervals.bed",
    repo_type="dataset"
)
df = pd.read_csv(file_path, sep='\t', header=None, names=['chrom', 'start', 'end'])