There are a lot of more interesting findings in this study. For more detail check out the preprint at medrxiv.org/content/10.1..., and feel free to DM me or drop me an email at a.xue@garvan.org.au if any questions. (15/n)

You are welcome to explore other TenK10K studies for different biological questions:
tinyurl.com/tenk10k-flag... led by
@annasecuomo.bsky.social
tinyurl.com/tenk10k-repeat led by
@htanudisastro.bsky.social
tinyurl.com/tenk10k-causal led by
@alberthenry.bsky.social & Anne Senabouth (14/n)

@htanudisastro.bsky.social, @zhenqiao.bsky.social and many others! (12/n)
A special shout out to the core contributors to the TenK10K cohort, Rachael McCloy, Venessa Chin, Katie de Lange, Gemma Figtree, Alex Hewitt, @dgmacarthur.bsky.social,
@drjosephpowell.bsky.social (13/n)

A big thanks to all co-authors, especially my supervisor
@drjosephpowell.bsky.social , & TenK10K team, Jayden Fan, Oscar Dong, @lawrencehuang.bsky.social , @petercallen.bsky.social
, Ellie Spenceley, Eszter Sagi-Zsigmond, @blakebowen.bsky.social, @annasecuomo.bsky.social, @alberthenry.bsky.social

.. using paired multiome data without QTL information. This improvement further enhanced gene regulatory network inference, leading to the identification of 128 additional transcription factor (TF)–target gene pairs (a 22% increase), some of which show druggable potential. (11/n)

scATAC-seq and caQTL signals also boost the gene regulatory network inference, especially when using unpaired multiome data. We inferred peak-to-gene relationships from unpaired multiome data by incorporating caQTL and eQTL, achieving up to 80% higher accuracy compared to (10/n)

The genetic impact on chromatin accessibility not only shows cell type-specific patterns but also varies across cell states. We further detected 3,080 caQTLs whose allelic effects showed significant interaction with epi-genetic age. (9/n)

Integrating caQTLs with GWAS+eQTL improves fine-mapping of causal variants. We pinpointed 671 credible sets for inflammatory bowel disease, 428 of which are single-variant sets, and replicated a causal variant for ETS2 in monocytes recently reported in Stankey et al. 2024. (8/n)

Next, we ask why do GWAS hits often miss eQTLs? We integrated 60 GWAS from disease and blood traits with eQTLs and caQTLs and found caQTL integration yields 9.8–30% more colocalizations than eQTLs alone, particularly at distal elements or loci with multiple causal variants. (7/n)

We highlight an interesting example where a chromatin peak chr10:45592479-45592785 shows a negative effect on the gene expression level of MARCH8 in NK cells but a positive effect in Conventional Dendritic Cell 2 (cDC2). (6/n)

Integrating caQTL results with eQTLs from scRNA-seq of 1,925 donors and 5.4M cells revealed over 70,000 colocalized signals, including 25,280 candidate cis-regulatory elements (cCREs) further supported by causal inference using Mendelian randomization (MR). (5/n)

More than half of caQTLs show cell type–specific patterns. For example, the chromatin peak chr13:24670806–24672096 contains caQTLs in CD4 TCM and CD14 monocytes, and their top variants (13:24671328:T:C in CD4 TCM and 13:24570579:C:A in CD14 Mono) are independent (LD ≈ 0). (4/n)

We curated one of the largest population-level (n = 1,042) scATAC-seq data from peripheral blood with WGS data, which enabled us to characterize 440,996 chromatin peaks across 28 immune cell types and mapped 243,273 caQTLs. (3/n)

Chromatin accessibility QTLs (caQTLs) directly capture the impact of non-coding variants on elements like enhancer and promoter, yet existing maps lack scale and diversity. Our study delivers a significant cell type–resolved caQTL resource in blood and demonstrates its translational utility. (2/n)