Arslan Zaidi

Effects of ancestry, agriculture, and lactase persistence on the stature of prehistoric Europeans https://www.biorxiv.org/content/10.1101/2025.07.11.664181v1

Short answer: twin estimates aren’t inherently wrong but they’re not the end all be all either.

Human geneticist here. I don’t think we have a gold standard. The estimands are different across approaches bc they make different assumptions esp about the true trait architecture. The true arch is unknown but the hope is diff estimates will converge once we rule out artifacts, confounding etc

last day tomorrow to register for a talk/poster at Midwest popgen. Please RT(? is that what they say here?)!

I probably won't have time to make a video out of this.

But the materials are available online, and should be pretty self-explanatory.

Hope this is useful to some people.

privefl.github.io/statgen-cour...

Maybe they’re not washing their hands thoroughly after not using water in the bathroom.

Thrilled to see the latest work from the lab (@genscapelab.bsky.social) out in Current Biology! Congratulations @jaurban2204.bsky.social ! 🎉

Grateful to have had the chance to contribute to this effort.

MWPG2025 Join us for the 10th Midwest Population Genetics conference, held at the University of Minnesota Twin Cities from August 21 - August 22, 2025.

Please register if you haven't already done so.
MidWest PopGen is happening in August at the University of Minnesota (Twin Cities)—abstract deadline for talk: July 11th, and poster: August 1st.

sites.google.com/umn.edu/mwpg...

SPC: a SPectral Component approach to address recent population structure in genomic analysis Population structure is a well-known confounder in statistical genetics, particularly in genome-wide association studies (GWAS), where it can lead to inflated test statistics and spurious associations...

📢 Just posted: Our preprint introducing SPC — Spectral Components — is now live on medRxiv!

Led by Dr. Ruhollah Shemirani and years in the making, this method offers a robust, scalable way to adjust for recent population structure in genomic analyses.
🔗 www.medrxiv.org/content/10.1...

1/9

Website and hotel info for Midwest popgen. Please register (free!) and book rooms asap to get ahead of the State Fair crowd.

The world may be aflame but we can try to find some community with each other. #mwpg25 slogan?

sites.google.com/umn.edu/mwpg...

Our new work is now out in AJHG "Natural selection acting on complex traits hampers the predictive accuracy of polygenic scores in ancient samples." Great collaboration with @delvecchyo.bsky.social and Emilia Huerta-Sanchez, led by Valeria Añorve-Garibay.

Midwest Population Genetics 2025 Registration

**10th Midwest PopGen Conference**
Location: UMN Twin Cities,
Date: Aug 21–22!
Events: Talks, Posters, dinner + MN State Fair.

Register: docs.google.com/forms/d/e/1F... -- Abstracts by June 15 // Register by Aug 1.

Add to calendar: calendar.google.com/calendar/eve...

#MidwestPopGen

The preliminary program of ProbGen 2026, which will be held at UC Berkeley, is now up: probgen2026.github.io

Sharing on behalf of Rasmus Nielsen who is not on this site. For more see his thread on that other site: x.com/ras_nielsen/...

Jennifer Blanc and @jeremyjberg.bsky.social use theory and simulations to study the process of testing for an association between polygenic scores and axes of ancestry variation when confounding factors are present.

Learn more about their findings in #GENETICS: buff.ly/EmKpXyP

When should adaptation arise from a polygenic response versus few large effect changes? https://www.biorxiv.org/content/10.1101/2025.05.15.654234v1

Our paper on the theoretical properties of SNP heritability in admixed populations is out in Genetics:
academic.oup.com/genetics/adv...

with @jinguohuang.bsky.social, @nicole-kleman.bsky.social , Saonli Basu, and Mark Shriver

Reminder to folks in genetics/genomics, HGG Advances is a wonderful open-access journal to publish your work in. Consider submitting your latest and greatest with us! :D

Out today in AJHG, this work from my (recently graduated) student Teng Heng on recessive effects in 44k British South Asians from the Genes & Health project @genesandhealth.bsky.social . She found 185 independent hits of which >40% were novel. Worth looking for these in your own cohorts!

I recently defended my PhD on "Mitochondrial DNA evolution alongside nuclear DNA".
Finishing my time at #PennState (with Kateryna Makova, @pennstateuniv.bsky.social),
as I transition to the #TwinCities (@mitopr.bsky.social @aazaidi.bsky.social)!

It’s not necessary. As long as it’s computed the same way for everyone you’re comparing, it doesn’t matter. Reason why you see those striped lines b/w UKB PRS and theirs in fig7 is because they were computed differently.

In the absence of other biases, random-effect estimators of h2snp are still interpretable: sum of h2 across loci --relevant for defining GWAS discoverability. But it is not the right quantity for defining the upper limit of PRS r2 where LD contribution matters.

Lots of other discussion in the paper, but an overall point is that h2snp in admixed and other structured populations can be biased for other reasons beyond just confounding.

This kind of bias has implications for GWAS cohorts where certain ancestries/communities might not be well-represented and is one of the few cases I've seen where it makes sense to remove individuals who may be 'outliers' in terms of genetic similarity from the rest of the cohort.

Another way I think about this is that the minor ancestry acts like a rare, localized environment (e.g. zaidi and mathieson elifesciences.org/articles/61548) that affects the distribution of summary statistics non-uniformly and therefore is not captured by the LDSC intercept.

However, if the `minor' ancestry is rare (first few generations of the CGF admixture model in the figure), cov-LDSC estimates are biased b/c of inflation in effect sizes in a way that is correlated with local levels of LD and not captured by the intercept.

HE regression performs similarly (with ancestry correction). cov-LDSC is more complicated. It gives similarly biased estimates to HE and GREML for the same reason (independent effect assumption) when ancestry components are well-represented (HI admixture model in the figure).

We showed that GREML can over- or under-estimate heritability even if causal variants are directly genotyped because it does not capture contributions of directional LD - which can be appreciable, especially for traits under selection - a consequence of its assumption that effects are independent.

Interpreting SNP heritability in admixed populations SNP heritability (h2snp) is defined as the proportion of phenotypic variance explained by genotyped SNPs and is believed to be a lower bound of heritability (h2), being equal to it if all causal varia...

An update to our preprint led by @jinguohuang.bsky.social and @nicole-kleman.bsky.social on interpreting SNP heritability in admixed populations, now with HE regression and LD Score regression in addition to GREML.

On ARGs, pedigrees, and genetic relatedness matrices Genetic relatedness is a central concept in genetics, underpinning studies of population and quantitative genetics in human, animal, and plant settings. It is typically stored as a genetic relatedness...

VERY excited to announce a pre-print that represents the culmination of several years of work*: “On ARGs, pedigrees, and genetic relatedness matrices”

doi.org/10.1101/2025...

Led by Brieuc Lehmann with contributions from @epigenci.bsky.social Luke Anderson Trocme Jerome Kelleher and Peter Ralph

My anxiety could never