Inflammatory bowel diseases such as Crohn’s disease and ulcerative colitis are incurable, often debilitating diseases which affect 1 in 123 people in the UK. Whilst drugs exist to treat them, for many these can be ineffective or lose efficacy over time…

…thus there is a need for new IBD drugs to be developed. As reported by @OpenTargets and others, drug targets with genetic support are far more likely to be approved for use in the clinic. We set out to create a dataset for finding novel IBD genes supported by genetics and…

…as a result generated a HUGE cross-tissue single-cell atlas using the three most relevant tissues for IBD: terminal ileum (most frequently inflamed in Crohn’s), rectum (most frequently inflamed in colitis) and IBD blood (enriched for immune cells often tricky to find in gut).

We used this atlas to map eQTLs (genome-wide association of genetic variants to gene expression) across the 86 cell types and 9 major populations in our atlas. With statistical colocalisation, we tested if our eQTLs were driven by the same genetic variants as IBD GWAS signals…

..and what we found was STRIKING! With our high resolution, IBD-relevant, single-cell data, our eQTLs colocalised with 74 (!) genomic regions where IBD genes couldn’t previously be found (more in @bradleyomics.bsky.social ’s thread as to why we think this might be). These genes included…

…MAML2 and ZMIZ1 (both key players in Notch signalling) - they each had an eQTL colocalise with an independent IBD GWAS locus, on chromosomes 11 and 10, respectively. And both showed the strongest evidence of colocalisation in very distinct cell types (T cells and macrophages) but…

…by leaning on the diversity of cell types in our atlas AND the genotype data paired with each cell type, we were able to probe where the lead variant’s effect was strongest and identify likely effector cell types for these Notch pathway genes, which were both strongest in cDCs!

Another story which emerged was that many eQTLs for genes involved in Wnt signalling colocalised with independent IBD GWAS loci too! These included MYC (a cancer gene), RNF14 (eQTL only seen in inflamed bowels), FUBP1 (binds MYC) amongst others. These eQTLs were observed…

…in and/or explained the most variance in epithelial cell types, offering a possible avenue to therapeutics which don’t target the immune system (unlike most IBD drugs). But the role of Wnt signalling in cancer necessitates a careful approach to avoid harmful side effects.

Last but by no means least, we chose to take a reversed approach and check which of our colocalised IBD genes were ALREADY drug targets for non-IBD diseases. With a little help from @chembl.bsky.social and @opentargets.org , we noticed two particularly interesting cases: PSEN2 and NDUFAF1…

…are both existing targets for gamma secretase inhibitors (Alzheimer’s) and mitochondrial complex I inhibitors/metformin (diabetes), respectively. But both drugs have had IBD-like side effects reported (diarrhoea, weight loss, abdominal discomfort/bleeding). GSIs never made it…

…past phase 3 clinical trials in-part due to the gastrointestinal symptoms that even included IBD in one individual! Our results could’ve identified these symptoms arising without the expensive clinical trials - human genetics can identify novel targets AND safety issues!

Whilst we’re incredibly excited about our biological findings, what excites us most is that this tissue-based, single-cell approach could be applied to other complex diseases to gain similar insights and unlock the hidden potential of GWAS!

Sadly this thread is long enough without me diving into genetics too but fortunately co-first author and eQTL superstar @bradleyomics.bsky.social has made an excellent thread covering our thoughts as to WHY we think sc-eQTL approaches are better-suited to GWAS: bsky.app/profile/brad...

I’ve loved working on this project (IBDverse) and these results wouldn’t have been possible without co-first author @bradleyomics.bsky.social the two supervisors @carlanderson.bsky.social and Tim Raine, as well as everyone in their teams, especially those who played a direct role processing…