Alex Strudwick Young

I am recruiting a quantitative/computational postdoc to my group at UCLA. This is a great opportunity to work on foundational theory, methods, and software in statistical genetics. Apply here: recruit.apo.ucla.edu/JPF10842. Please repost!

Herasight does genetic testing of embryos, but we keep discovering people’s unknown medical conditions as a side-effect of our work. A few months ago we sequenced the genome of one of our employees and discovered he had an unusually high risk of psoriasis. He had a higher risk than 97% of the general population, largely thanks to the presence of HLA-C*06:02/PSORS1. For years, his mother had been suffering from a mysterious illness causing increasingly severe neck and back pain. It would often take her an hour or more to get out of bed in the morning. She had seen doctors multiple times over the years and described her symptoms, but they chalked it up to age-related mechanical back pain. She was told to get a different pillow and to replace her mattress. None of it helped. Her symptoms continued to worsen and her daily dose of pain relievers continued to climb. When our employee received his genetic risk report, he immediately began to wonder whether his high risk might be connected to her symptoms. Sure enough, when we sequenced her genome, his suspicions were confirmed. Her risk was higher than that of 99.3% of the general population! She immediately visited a rheumatologist and was diagnosed with psoriatic arthritis! The doctor put her on a DMARD, which has finally improved her symptoms. But what about his children? Will they inherit his high risk of this disease, just like he did from his mother? Fortunately, the answer is probably not! When we looked at his embryos we found not only do some have lower risk than him, one actually has a risk even lower than average! If you’ve got a family history of some disease you want to avoid passing on to your kids, please reach out! We can often help you lower that risk for your children. cal.com/team/herasight…

777 studies, 745 with summary statistics - the Genes & Health WES study is now in the GWAS Catalog, aiming to improve health outcomes for South Asian populations. Go explore 👇 ebi.ac.uk/gwas/publicati…

Interested in ancient environmental DNA, metagenomics, and bioinformatics? Sign up for our PhD course in Ancient Environmental Genomics at the University of Copenhagen, 20 to 25 Sept 2026. Sign-up deadline: 1 Aug 2026 phdcourses.ku.dk/DetailKursus.a…

@unboxpolitics @CarlosEAlvare17 @cremieuxrecueil @RegimeEnforcer Yeah but multiple testing correction is tricky when you have polygenic scores with very different levels of power.

@AlexTISYoung @CarlosEAlvare17 @cremieuxrecueil I think @RegimeEnforcer made a fair point that the within-family educational attainment results probably wouldn’t survive multiple testing corrections. Of course, this just goes to your point that more powerful analyses would be helpful to validate the results.

Family-GWAS results were critical in testing the robustness of inferences of natural selection in the recent Akbari/Reich paper. We're releasing family-GWAS sumstats on 28 phenotypes on thessgac.com data portal, using the powerful method we developed in Guan et al.

The family gwas results in akbari also support selection on bmi and ever-smoker. Height signal is driven by pop. strat as they acknowledge. Family-gwas for intelligence is small so the analysis likely underpowered. EA has the most consistent signal from the family gwas analyst across the three selection tests so they highlight that. But overall we need more powerful family-gwas across lots of different traits to really powerfully test these results.

actually I am not sure i believe this? why would the signal you think would apriori be the weakest(education by years of it) (ancestral environment did not have this, why would it even be a proxy for something else it picked up on, which naturally would be intelligence but intelligence itself does not survive) it seems to me that almost any other trait I can see a reason for selection on(height, disease related immune things, health related things..)

@elephant_ben The authors say that themselves

@elephant_ben Well height is a good counterexample. It shows signal using standard GWAS but not family-GWAS so the signal is likely driven by pop. strat.

Is Alexis T Young skeptical of Reich's latest paper? What an oddly subliminal tweet.

@elephant_ben A lot of the other family-GWAS are underpowered for this

Seems his view is that the Edu attainment going through means it isn't effected by pop strat but then what about all the other ones that didn't work? x.com/AlexTISYoung/s…

@unboxpolitics I don't think so. This method only works in weakly structured populations. The key issue is imputing in an unbiased way.

@AlexTISYoung Very cool! Is it possible to generalize this strategy to family-based admixture studies? In other words, linearly impute the ancestry proportions of the parents of singletons in an unbiased manner to increase power of a family-based admixture study?

Preprint from @florajay_ using attention neural networs to infer population size. Read the full preprint here: biorxiv.org/content/10.648…

@CarlosEAlvare17 @cremieuxrecueil They used the results of my family gwas study to test the polygenic signals of selection. They got significant results for educational attainment which is a good argument for the result not being pop strat. I've been meaning to write more about this but I've been too busy

Thanks, Alex. Could you clarify something on a claim that’s out there? Is it true that all the evidence for positive evolutionary selection for intelligence in Akbari… Reich is coming from GWAS PGS signal? If so, that merits the caveat that it could be false signal due to population structure and requires validation in family studies, right? How does this family GWAS relate to such proof?

@lu_sichu Often ideas that are so simple they're almost dumb are the best

banger

In response to demand we've released an expanded set of sumstats from applying the unified estimator to 28 phenotypes in the UK Biobank, with sumstats available for ~5 million high quality imputed SNPs (INFO>0.99, MAF>1%).

Empirically this up to doubles the effective sample size for estimation of direct genetic effects in UK Biobank compared to sib-GWAS.

Can we predict high myopia risk at birth? 👁️🧬 Our new @NatureGenetics paper leverages >1.7M people to decode the genetic basis of myopia. We trained a powerful risk prediction model—an early "warning radar" to protect vision. Proud of the team! nature.com/articles/s4158…