Chromatin-level QTLs in immune cells recover 4.5–22.6% more disease signals than expression QTLs alone

Genetic effects on chromatin accessibility can now be mapped within 28 immune cell types rather than inferred indirectly through gene expression. Most previous molecular QTL maps came from bulk RNA sequencing eQTLs, which have low sensitivity for cell-type-specific regulatory effects, leaving many genome-wide association signals without a regulatory explanation. In 922 donors with matched whole-genome sequencing and 3,472,552 quality-controlled nuclei, 100,483 of 440,996 peaks carried at least one significant association, giving 243,225 caQTLs, 60% of them cell-type specific.

Accessibility-level associations recover disease signals that expression-level associations miss, and the gain comes mainly from cell-type resolution and distal regulation. Colocalization with expression QTLs alone left many non-coding risk loci without a regulatory interpretation. For 16 diseases and 44 blood traits, caQTLs colocalized with 4.5–22.6% more signals than eQTLs alone; caQTLs retained an average 15.9% advantage even against eQTLs widened to ±1 Mb, and the share of inflammatory bowel disease loci with colocalization evidence rose from about 16% to about 65%.

The same data can improve regulatory network inference by using genetic variants as causal anchors between peaks and genes. Earlier peak–gene linking methods relied largely on co-expression or co-accessibility and did not use genetic variation. On the 10X Genomics PBMC 10K multiome dataset, adding Mendelian randomization results from caQTLs and eQTLs to graph neural network inference produced 128 additional transcription factor–target pairs in CD14 monocytes, a 22.4% increase, 66 of them at GWAS loci for the 16 disease traits.

The results are demonstrated in peripheral blood mononuclear cells from resting samples at a 28-cell-type annotation granularity, and peaks vary with scATAC-seq platform and peak-calling strategy; the authors state that caQTL mapping should extend to disease-relevant tissues and incorporate histone modifications, DNA methylation and protein expression.

The authors note that caQTL power is sensitive to the cis-window, with narrow windows missing distal effects and wide windows requiring control of technical noise; cell-type annotation lacks a gold standard and rare populations are limited by nuclei per donor; the remaining colocalization gap between caQTLs and eQTLs is attributed to mechanisms not tested directly, such as post-transcriptional buffering.