Datasets:

Genentech
/

microglia-scatac-tutorial-data

license: mit
task_categories:
  - tabular-regression
tags:
  - biology
  - genomics
pretty_name: gReLU tutorial 3 dataset (Microglia scATAC-seq)
size_categories:
  - 10K<n<100K

microglia-scatac-tutorial-data

Dataset Summary

This dataset contains pseudobulk scATAC-seq data for human microglia, derived from the study by Corces et al. (2020) (https://www.nature.com/articles/s41588-020-00721-x). Genome coordinates correspond to the hg38 reference genome. This data is used in tutorial 3 of gReLU (https://github.com/Genentech/gReLU/blob/main/docs/tutorials/3_train.ipynb).

Dataset Structure

The dataset is divided into two files: peaks and fragments.

Statistics

File	Rows	Description
`peak_file.narrowPeak`	83,320	Called ATAC-seq peaks
`fragment_file.bed`	57,919,016	Individual ATAC-seq fragments

1. Peaks file (`peak_file.narrowPeak`)

Standard ENCODE narrowPeak format (tab-separated, no header).

chrom: Chromosome / Contig name
start: 0-based start position
end: End position
name: Peak identifier
score: Integer score for display
strand: Orientation
signalValue: Measurement of overall enrichment
pValue: Statistical significance (-log10)
qValue: False discovery rate (-log10)
peak: Point-source (summit) relative to start

2. Fragments file (`fragment_file.bed`)

Standard BED6 format representing individual ATAC-seq fragments.

chrom: Chromosome
start: Start position
end: End position
source: Sequencing run identifier (e.g., SRR11442505)
score: Placeholder (0)
strand: Orientation

Usage

from huggingface_hub import hf_hub_download
import pandas as pd

peak_path = hf_hub_download(
    repo_id="Genentech/microglia-scatac-tutorial-data",
    repo_type="dataset",
    filename="peak_file.narrowPeak"
)
peaks = pd.read_csv(peak_path, sep='\t', header=None)

frag_path = hf_hub_download(
    repo_id="Genentech/microglia-scatac-tutorial-data",
    repo_type="dataset",
    filename="fragment_file.bed"
)
fragments = pd.read_csv(frag_path, sep='\t', header=None,
                        names=['chrom', 'start', 'end', 'source', 'score', 'strand'])