Genetics: Chromosome 17 CA10

[ME/CFS Science Blog]

I did the same analysis as for the chr20 region. Finemapping with Susie gave a credible set with 64 SNPs that is hard to reduce to a smaller number: the probability is spread out across these SNPs. CA10 is the only protein-coding gene nearby but a lncRNA called ENSG00000310017 falls inside our SNP region but this isn't where the highest probability lies.




The Encode chromatin data from SCREEN includes 10 regulatory elements with at least one of the 64 SNPs from our ME/CFS credible set. And these 10 are mostly exclusively open in brain tissue with top hits for the cerebellar cortex and spinal cord (although caveat: there were only a limited number of samples in these categories compared to cortex or other tissue). I’ve tried to group them into my own-made categories in the plot below. On the left you see the 10 cCREs (candidate cis-regulatory elements) with the SNPs that fall inside them. Higher z-values means more regulatory activity.



I also looked at the Li 2023 brain atlas which differentiates per cell type in the brain. The ME/CFS SNPs were mostly open in glutamergic Intratelencephalic (cortical layer 2 to 6) cell types.



Here’s another plot that puts a lot of info together.

 

[hotblack]

Interesting stuff @ME/CFS Science Blog Would you say this supports the previous idea of this being promoters or enhancers for CA10?

 

[tralfamadorian97]

Finemapping with Susie gave a credible set with 64 SNPs that is hard to reduce to a smaller number: the probability is spread out across these SNPs

Nice. A few weeks ago I tried adding various Bayesian priors to SUSIE as described in this paper (the method is called PolyFun). PolyFun overweights genetic variants that have S-LDSC annotations associated with a large amount of heritability. The first PIP plot below uses a uniform prior (i.e. standard SUSIE, which should be similar to your results). The second and third PIP plots use different versions of the PolyFun prior. They upweight a few variants annotated as "evolutionarily conserved". The logic is that variants conserved by evolution are typically important, and so tend to carry disproportionate heritability in disease GWAS.

The PolyFun paper provides experiments indicating that on average over many loci, use of this kind of annotation-based prior improves fine-mapping accuracy. However, there is no guarantee that a PolyFun prior is better than the uniform prior on any individual locus.





I wrote up the results here.

 

[ME/CFS Science Blog]

Would you say this supports the previous idea of this being promoters or enhancers for CA10?

I don't think the chromatin data gives answers to this question but I'll analyse further using phewas and eQTL and will try to get back to this.