Summary
Researchers report a single-nucleus atlas spanning 21 adult human tissues and four donors, combining gene-expression and chromatin-accessibility data. It maps candidate regulatory elements and gene links, and uses sequence-to-function models to predict effects of fine-mapped variants.
A bioRxiv preprint presents a gene-regulation atlas built from 459,856 transcriptomic and chromatin-accessibility profiles across 21 adult human tissues and four donors. The researchers use the combined data to map cell types, candidate regulatory DNA and links between regulatory elements and genes, then train models to predict how fine-mapped genetic variants may affect chromatin accessibility.
The atlas includes paired measurements from 160,688 nuclei. The work is a reference map of gene regulation at single-nucleus resolution, rather than a clinical study; its variant-effect results are computational predictions. The paper is a preprint, so these findings are presented before peer review.
A joint map of gene activity and accessible DNA
The two measurements capture different parts of gene regulation. Transcriptomic profiling records which genes are being expressed in a nucleus. Chromatin-accessibility profiling identifies DNA regions that are open to regulatory machinery. Measuring both in the same nucleus can help connect a regulatory region with the genes active in that cell, while measurements from other nuclei extend the map across tissues and cell populations.
The atlas resolves nine cell lineages, 61 broad cell types and 313 subclusters. The researchers identify 1,085,062 candidate cis-regulatory elements, or cCREs: DNA regions that may help control nearby or distant genes. Of these, 161,270 are reported as absent from ENCODE, an existing reference resource for functional genomic elements.
Joint profiling yielded 871,177 cCRE–gene associations. These candidate links offer a way to investigate how regulatory DNA relates to gene expression in particular cell types. The researchers report that regulatory activity was driven mainly by cell identity, with tissue context adding further differences. They also identify lineage-restricted and constitutively inaccessible chromatin domains, and report preferential hypomethylation of the latter across human cancers.
Modelling the effects of genetic variants
The team used the atlas to train sequence-to-function models, which estimate how DNA sequence changes may affect a measurable molecular feature. The models predicted chromatin-accessibility effects for 548,656 fine-mapped variants and classified 18,133 as high-effect variants, including 1,120 described as broadly active.
These predictions connect genetic variants to possible changes in regulatory activity, a step that can help researchers investigate how non-coding genetic risk relates to particular cells and tissues. The authors also trained models for eight endothelial subtypes, allowing predicted effects to be considered across different vascular beds.
The atlas draws on four donors, so it provides a detailed cellular reference rather than a population-wide account of human variation. Its predicted variant effects are computational results that can help prioritise questions for further study; they are not direct measurements of each variant’s effect in a person.