Hugging Face's logo

Datasets:
anglixue
/
TenK10K_multiome

Languages:
English
Size:
n>1T
Tags:
single-cell
multiome
caQTL
eQTL
colocalization
SMR
License:
Dataset card Files
xet
 Community
The Dataset Viewer has been disabled on this dataset.

⚠️ Warning:
This dataset contains summary statistics derived from a preprint that has not yet undergone full peer review and may be subject to change prior to formal publication. Any use of this dataset should therefore be interpreted with caution.

Dataset Card for Dataset Name

This dataset card aims to be a base template for new datasets. It has been generated using this raw template.

Dataset Details

Dataset Description

  • Curated by: TenK10K multiome team
  • Funded by: NHMRC
  • Shared by: Angli Xue
  • Language(s) (NLP): English
  • License: cc-by-4.0

This dataset provides population-scale, single-cell resolution summary statistics of genetic effects on chromatin accessibility in human peripheral blood mononuclear cells (PBMCs), generated as part of the TenK10K program. It includes chromatin accessibility QTL (caQTL) summary statistics mapped across 28 immune cell types from 922 donors, alongside colocalization results, fine-mapping outputs, ChromBPNet variant effect scores, cell state-dependent caQTL results, and gene regulatory network inference outputs.

Dataset Sources

Uses

Direct Use

This dataset is intended for researchers studying the genetic regulation of gene expression and chromatin accessibility in human immune cells. Suitable use cases include but not limited to:

  • Identifying regulatory mechanisms underlying GWAS loci for complex diseases and blood traits by integrating caQTL summary statistics with external GWAS data
  • Linking chromatin peaks to target genes using the provided colocalization, SMR, and GLUE peak–gene results
  • Inferring transcription factor–target gene regulatory networks in specific immune cell types
  • Investigating cell state-dependent genetic effects on chromatin accessibility using the epigenetic age interaction results
  • Benchmarking or training deep learning models for regulatory variant effect prediction using the ChromBPNet weights and scores
  • Fine-mapping causal variants jointly affecting chromatin accessibility, gene expression, and disease risk using the provided mvSuSiE credible sets

This dataset is not intended for clinical decision-making or individual-level genetic inference, as only population-level summary statistics are released.

Out-of-Scope Use

This dataset mush NOT be used for de-identifying individual-level information

Dataset Structure

  • Peak annotations: Functional annotations of 440,996 consensus chromatin peaks across 28 immune cell types
  • caQTL summary statistics: Common and rare variant associations with chromatin accessibility per cell type
  • Fine-mapping results: SuSiE and mvSuSiE credible sets for causal variant identification
  • ChromBPNet model weights and predicted scores: Deep learning-based variant effect predictions
  • Cell state-dependent caQTL summary statistics: Genotype-by-epigenetic-age interaction results
  • Colocalization results: coloc and SMR outputs for caQTL–eQTL, caQTL–GWAS, and eQTL-GWAS colocalization
  • GLUE peak–gene and TF–gene links: Inferred cis-regulatory and gene regulatory network relationships

Dataset Creation

Curation Rationale

This dataset was created to provide a large-scale, single-cell resolution map of genetic effects on chromatin accessibility in human peripheral blood mononuclear cells (PBMCs), as part of the TenK10K program. It addresses the limited overlap between GWAS loci and eQTLs by mapping chromatin accessibility QTLs (caQTLs) across 28 immune cell types, enabling more comprehensive dissection of non-coding genetic variation underlying complex diseases.

Source Data

Data Collection and Processing

Single-cell ATAC-seq data were generated from 952 donors (Tasmanian Ophthalmic Biobank/OneK1K cohort) using the 10x Genomics scATAC-seq v2 kit, with multiome (RNA+ATAC) data from 90 additional donors (BioHeart and LBIO cohorts). Whole-genome sequencing was performed at 30x coverage. After quality control, 3.47 million nuclei from 922 donors were retained. Chromatin peaks were called per cell type using MACS3 via SnapATAC2 and merged into 440,996 consensus peaks. caQTL mapping used TensorQTL with permutation-based correction across a ±1 Mb cis-window. Colocalization was performed with coloc and SMR; gene regulatory network inference used GLUE; fine-mapping used SuSiE and mvSuSiE; and ChromBPNet was used for deep learning-based variant effect prediction.

Who are the source data producers?

Data were generated by researchers at the Garvan Institute of Medical Research and the University of New South Wales, Sydney, Australia, as part of the TenK10K study. Participants are adults recruited through the Tasmanian Ophthalmic Biobank, BioHEART, and Liquid Biopsy Biobank cohorts, all of whom provided informed consent under approved ethics protocols.

Personal and Sensitive Information

This dataset contains human genomic data. All individual-level genotype and single-cell data have been removed; only aggregated summary statistics are included in this release. Data were collected under ethics approval (Tasmanian Ophthalmic Biobank: 2020/ETH02479; BioHEART: 2019/ETH08376; Liquid Biopsy Biobank: 2019/ETH13113), and all participants provided informed consent.

Bias, Risks, and Limitations

  • Ancestry bias: The cohort is predominantly of European ancestry, which may limit the generalisability of caQTL findings to other populations.
  • Cell type power imbalance: caQTL discovery power varies substantially across cell types depending on cell abundance; rare cell types (e.g., plasmablasts, HSPCs) are likely underpowered and may have higher false negative rates.
  • Tissue specificity: This release covers only peripheral blood immune cells; regulatory effects identified here may not generalise to other tissues or disease-relevant cell types not present in circulation.
  • Summary statistics only: Individual-level data are not released; some analyses requiring raw data (e.g., novel QTL mapping with different parameters) cannot be reproduced from this release alone.
  • Peak-based representation: caQTLs are mapped to peaks, which are not fixed genomic features and may vary across studies depending on sequencing platform and peak-calling strategy, complicating direct cross-study comparisons.

Recommendations

  • Results should be interpreted in the context of a predominantly European cohort; replication in diverse ancestries is strongly encouraged before drawing population-level conclusions.
  • When integrating with external GWAS or eQTL datasets, users should ensure consistent genome build (GRCh38) and variant annotation conventions.
  • caQTL signals from low-abundance cell types should be treated with caution and ideally validated in independent datasets.
  • This dataset is intended for research purposes only and should not be used for clinical decision-making or individual genetic risk assessment.
  • Users replicating or extending these analyses are encouraged to consult the accompanying code repository at https://github.com/powellgenomicslab/tenk10k_phase1_multiome.

Citation

BibTeX:

@article{xue2025multiome, author = {Xue, Angli and others}, title = {Genetic regulation of cell type--specific chromatin accessibility shapes immune function and disease risk}, journal = {medRxiv}, year = {2025}, doi = {10.1101/2025.08.27.25334533}, url = {https://www.medrxiv.org/content/10.1101/2025.08.27.25334533v2}, note = {Preprint} }

APA:

Xue, A., Fan, J., Dong, O. A., Huang, H. L., Chen, L., Allen, P. C., Spenceley, E., ... & Powell, J. E. (2025). Genetic regulation of cell type–specific chromatin accessibility shapes immune function and disease risk. medRxiv, 2025-08.

Glossary

  • caQTL (Chromatin Accessibility QTL): A genetic variant associated with differences in chromatin accessibility at a specific genomic region across individuals.
  • caPeak: A chromatin accessibility peak that has at least one significant caQTL association.
  • eQTL (Expression QTL): A genetic variant associated with differences in gene expression levels across individuals.
  • scATAC-seq: Single-cell Assay for Transposase-Accessible Chromatin with sequencing; measures chromatin accessibility at single-cell resolution.
  • Multiome: A sequencing assay that simultaneously measures gene expression (RNA) and chromatin accessibility (ATAC) from the same cell.
  • coloc: A Bayesian statistical method to test whether two traits (e.g., a caQTL and a GWAS signal) share a common causal variant at a locus.
  • SMR (Summary-based Mendelian Randomisation): A method that uses QTL summary statistics to infer causal relationships between molecular phenotypes and complex traits.
  • PP.H4: Posterior probability that two traits share a single causal variant, as estimated by coloc.
  • SuSiE / mvSuSiE: Sum of Single Effects model for fine-mapping causal variants, in univariate and multivariate settings respectively.
  • ChromBPNet: A deep learning model trained on scATAC-seq data to predict the regulatory effects of genetic variants at base resolution.
  • GLUE: Graph-Linked Unified Embedding; a multi-omics integration framework used here to infer peak–gene and TF–gene regulatory relationships.
  • GRN (Gene Regulatory Network): A network describing regulatory relationships between transcription factors and their target genes.
  • PBMC (Peripheral Blood Mononuclear Cell): A blood cell with a round nucleus, including lymphocytes and monocytes, commonly used in immune profiling studies.
  • Epigenetic age: A relative measure of cellular maturation inferred from chromatin accessibility patterns at age-associated genomic regions.

More Information

Dataset Card Authors

Angli Xue, Garvan Institute of Medical Research, Australia

Key contributors to the dataset: Angli Xue, Jianan Fan, Oscar A. Dong, Hao Lawrence Huang, Chen Ling, Peter C. Allen, Eleanor Spenceley, Blake Bowen

Dataset Card Contact

a.xue@garvan.org.au
Downloads last month
7,492
Total file size:
1.3 TB