Kaur Alasoo
@kauralasoo.bsky.social
Associate Professor of Bioinformatics at University of Tartu, Estonia. Project lead at eQTL Catalogue. https://kauralasoo.github.io/
This is one of the coolest studies that I have seen recently! Thanks for highlighting, @jeffreypullin.bsky.social!
Genes and environment profoundly affect the human virome
Genes and environment profoundly affect the human virome
October 2, 2025 at 4:42 PM
This is one of the coolest studies that I have seen recently! Thanks for highlighting, @jeffreypullin.bsky.social!
Genes and environment profoundly affect the human virome
Genes and environment profoundly affect the human virome
Thanks, I had missed that GWAS! We expored the SP110/SP140 locus a bit, but ultimately decided not to focus on it, because it had high heterogeneity between cohorts and did not replicate in MAGE.
October 2, 2025 at 4:25 PM
Thanks, I had missed that GWAS! We expored the SP110/SP140 locus a bit, but ultimately decided not to focus on it, because it had high heterogeneity between cohorts and did not replicate in MAGE.
A big shout out to the @finngen.bsky.social team for releasing their FinnGen+MVP+UKBB meta-analysis ahead of publication at the end of last year (mvp-ukbb.finngen.fi). This is an amazing resources that allowed us to replicate our USP18 trans-eQTL and lupus colocalisation in an independent cohort:
October 2, 2025 at 1:05 PM
A big shout out to the @finngen.bsky.social team for releasing their FinnGen+MVP+UKBB meta-analysis ahead of publication at the end of last year (mvp-ukbb.finngen.fi). This is an amazing resources that allowed us to replicate our USP18 trans-eQTL and lupus colocalisation in an independent cohort:
And as a badge of honour, I can now say that I have authored a paper that has been rejected from Scientific Reports for lack of novelty!
September 27, 2025 at 8:53 AM
And as a badge of honour, I can now say that I have authored a paper that has been rejected from Scientific Reports for lack of novelty!
Perhaps the biggest change is that we have added some practical guidelines for performing MR with proxy exposures. These are summarised on this nice figure prepared by Ida Rahu:
September 27, 2025 at 8:53 AM
Perhaps the biggest change is that we have added some practical guidelines for performing MR with proxy exposures. These are summarised on this nice figure prepared by Ida Rahu:
We did see that weaker and more distal pQTL signals (e.g. 4th or 5th credible sets) had lower precision than the stronger (and more proximal ones).
June 9, 2025 at 8:28 PM
We did see that weaker and more distal pQTL signals (e.g. 4th or 5th credible sets) had lower precision than the stronger (and more proximal ones).
But there will be plenty of other cases where single causal gene (out of multiple co-regulated ones) is much more likely. My favorite example is the fine-mapped eQTL for SORT1 ( rs12740374, PIP = 0.99) that colocalises with CAD and LDL cholesterol.
June 9, 2025 at 8:18 PM
But there will be plenty of other cases where single causal gene (out of multiple co-regulated ones) is much more likely. My favorite example is the fine-mapped eQTL for SORT1 ( rs12740374, PIP = 0.99) that colocalises with CAD and LDL cholesterol.
Hard to say. I think this depends on which genes are at the same locus. For example, there is a GWAS hit for lupus at the IFNa/b cluster and it is highly likely that any regulatory variant at this locus would act via multiple genes.
June 9, 2025 at 8:07 PM
Hard to say. I think this depends on which genes are at the same locus. For example, there is a GWAS hit for lupus at the IFNa/b cluster and it is highly likely that any regulatory variant at this locus would act via multiple genes.
Coming to #eshg2025 #eshg25? Check out these talks and poster from our team!
@genomicsdoge.bsky.social @estbiobank.bsky.social @peepkolberg.bsky.social
@genomicsdoge.bsky.social @estbiobank.bsky.social @peepkolberg.bsky.social
May 24, 2025 at 7:38 AM
Coming to #eshg2025 #eshg25? Check out these talks and poster from our team!
@genomicsdoge.bsky.social @estbiobank.bsky.social @peepkolberg.bsky.social
@genomicsdoge.bsky.social @estbiobank.bsky.social @peepkolberg.bsky.social
In all three gene regions we are seeing a null effect on T2D (and CAD).
April 13, 2025 at 7:48 AM
In all three gene regions we are seeing a null effect on T2D (and CAD).
Here's the putative causal DAG. Instead of estimating the genome-wide average effect of BCAAs on T2D (which could be confounded by many factors), we are now asking if altered protein function of BCAT2/DBT/PPM1K might have a causal effect on T2D risk.
April 13, 2025 at 7:48 AM
Here's the putative causal DAG. Instead of estimating the genome-wide average effect of BCAAs on T2D (which could be confounded by many factors), we are now asking if altered protein function of BCAT2/DBT/PPM1K might have a causal effect on T2D risk.
We also performed genome-wide Mendelian randomisation to explore links between metabolic traits and CAD/T2D. While this did recover known causal effects of LDL cholesterol on CAD and glucose on T2D, almost everything else appeared to be significant as well.
April 13, 2025 at 7:48 AM
We also performed genome-wide Mendelian randomisation to explore links between metabolic traits and CAD/T2D. While this did recover known causal effects of LDL cholesterol on CAD and glucose on T2D, almost everything else appeared to be significant as well.
Using @genomicsengland.bsky.social imputation allowed us to test many more low-frequency variants in the UK Biobank than previous studies (95M total). As an example, in branched-chain amino acid GWAS, we detected hits for all five members of the BCKDH complex, two of which had MAF < 0.0005!
April 13, 2025 at 7:48 AM
Using @genomicsengland.bsky.social imputation allowed us to test many more low-frequency variants in the UK Biobank than previous studies (95M total). As an example, in branched-chain amino acid GWAS, we detected hits for all five members of the BCKDH complex, two of which had MAF < 0.0005!
We've just posted an updated version of our metabolic trait GWAS preprint. The GWAS summary statistics are still the same as in October, but we've included several new follow-up analyses and completely rewritten the manuscript! See thread below for some highlights.
www.medrxiv.org/content/10.1...
www.medrxiv.org/content/10.1...
April 13, 2025 at 7:48 AM
We've just posted an updated version of our metabolic trait GWAS preprint. The GWAS summary statistics are still the same as in October, but we've included several new follow-up analyses and completely rewritten the manuscript! See thread below for some highlights.
www.medrxiv.org/content/10.1...
www.medrxiv.org/content/10.1...
I think that in your example the causal variant is probably somewhere near the UGT1A1 promoter (where the eQTL variants with the smallest p-value are) and the UGT1A4 missense variant just happens to be in LD with it and might not do anything. This is supported by EBI ProtVar:
March 4, 2025 at 8:45 PM
I think that in your example the causal variant is probably somewhere near the UGT1A1 promoter (where the eQTL variants with the smallest p-value are) and the UGT1A4 missense variant just happens to be in LD with it and might not do anything. This is supported by EBI ProtVar:
And the same variant is also a fine mapped sQTL (PIP ~ 0.99 in multiple studies) in the @eqtlcatalogue.bsky.social
elixir.ut.ee/eqtl/?rsid=r...
elixir.ut.ee/eqtl/?rsid=r...
March 4, 2025 at 7:55 PM
And the same variant is also a fine mapped sQTL (PIP ~ 0.99 in multiple studies) in the @eqtlcatalogue.bsky.social
elixir.ut.ee/eqtl/?rsid=r...
elixir.ut.ee/eqtl/?rsid=r...
The variant is 2bp from a splice junction and predicted to cause to splice donor loss by both SpliceAI and Pangolin.
spliceailookup.broadinstitute.org#variant=chr1...
spliceailookup.broadinstitute.org#variant=chr1...
March 4, 2025 at 7:52 PM
The variant is 2bp from a splice junction and predicted to cause to splice donor loss by both SpliceAI and Pangolin.
spliceailookup.broadinstitute.org#variant=chr1...
spliceailookup.broadinstitute.org#variant=chr1...
The missense effect of the rs117048185 is predicted to be benign by most models.
March 4, 2025 at 7:50 PM
The missense effect of the rs117048185 is predicted to be benign by most models.
Inspired by @ericfauman.bsky.social, here's a slightly different GWAS riddle. The trait is valine (plasma concentration) and the lead variant is rs117048185.
What is the causal gene?
What is the mode of action of the causal variant (expression altering, splice altering, missense, etc)?
Hint:
What is the causal gene?
What is the mode of action of the causal variant (expression altering, splice altering, missense, etc)?
Hint:
March 4, 2025 at 5:27 PM
Inspired by @ericfauman.bsky.social, here's a slightly different GWAS riddle. The trait is valine (plasma concentration) and the lead variant is rs117048185.
What is the causal gene?
What is the mode of action of the causal variant (expression altering, splice altering, missense, etc)?
Hint:
What is the causal gene?
What is the mode of action of the causal variant (expression altering, splice altering, missense, etc)?
Hint:
eQTL Catalogue lists three fine mapped signals for CHRNA2 in GTEx cerebellum, but rs11783093 does not seem to be any of them. It is very nice to see though that the primary CHRNA2 cerebellum eQTL signal (elixir.ut.ee/eqtl/?credib...) is in perfect LD with another smoking initiation lead variant.
February 22, 2025 at 9:56 AM
eQTL Catalogue lists three fine mapped signals for CHRNA2 in GTEx cerebellum, but rs11783093 does not seem to be any of them. It is very nice to see though that the primary CHRNA2 cerebellum eQTL signal (elixir.ut.ee/eqtl/?credib...) is in perfect LD with another smoking initiation lead variant.
Oh, and @estbiobank.bsky.social does an absolutely great job communicating CAD risk to participants.
February 7, 2025 at 12:11 PM
Oh, and @estbiobank.bsky.social does an absolutely great job communicating CAD risk to participants.
I agree. Regenie even gives you -log10p automatically. Unfortunately we still had to commit this sin when setting up a pheweb for our latest metabolic biomarker GWAS, because pheweb does not support -log10 p-values.
All our top 1000 hits have p < 1e-300. nmrmeta.gi.ut.ee/top_hits
All our top 1000 hits have p < 1e-300. nmrmeta.gi.ut.ee/top_hits
January 20, 2025 at 4:21 PM
I agree. Regenie even gives you -log10p automatically. Unfortunately we still had to commit this sin when setting up a pheweb for our latest metabolic biomarker GWAS, because pheweb does not support -log10 p-values.
All our top 1000 hits have p < 1e-300. nmrmeta.gi.ut.ee/top_hits
All our top 1000 hits have p < 1e-300. nmrmeta.gi.ut.ee/top_hits
I would probably start with BLUEPRINT (QTS000002). These are the eGene counts (FDR < 0.05 I believe) from some old analysis:
January 3, 2025 at 8:56 AM
I would probably start with BLUEPRINT (QTS000002). These are the eGene counts (FDR < 0.05 I believe) from some old analysis:
We did not see many pleiotropic effects, but this might just reflect low power due to low AF as well as relatively small effect of the missense variant on protein function (glucose beta = -0.072). Here's the PheWAS plot from the @opentargets.org genetics portal:
January 3, 2025 at 8:52 AM
We did not see many pleiotropic effects, but this might just reflect low power due to low AF as well as relatively small effect of the missense variant on protein function (glucose beta = -0.072). Here's the PheWAS plot from the @opentargets.org genetics portal: