sgravel.bsky.social

An ever-slightly non-PSD matrix allows your favourite algorithm to find loophole states with negative errors. Ulises shows how to efficiently restore the PSD condition with no cost to algorithm performance, little to no approximation cost, and a large potential bonus to stability.

Because LD matrices are large, efficient methods take shortcuts, for example by neglecting LD between distant SNPS. These shortcuts seem benign, but sometimes cause the LD matrix to lose a mathematical property called positive semi-definiteness (PSD). This is a problem.

LD Matrix Approximations for Scalable Analysis of High-dimensional Genetic Data Linkage disequilibrium (LD) matrices are an essential part of many statistical genetics methods. However, their high dimensionality makes their computation and storage impractical for large genomic da...

To analyze millions of genetic variants, we need tools that are fast and memory efficient. Some of the most efficient tools have problems with numerical instability on dense sets of snps. Bercovich et al. figured out how to overcome a common source of instability: www.biorxiv.org/content/10.1...

Sex differences in transcription-associated mutagenesis in the human germline https://www.biorxiv.org/content/10.1101/2025.09.18.677082v1

Multiethnic polygenic risk scores improve risk prediction in diverse populations - PubMed Methods for genetic risk prediction have been widely investigated in recent years. However, most available training data involves European samples, and it is currently unclear how to accurately predic...

Finally, it shows that integrating multiple sources of heterogeneity does not need to be complicated. The work was inspired by excellent previous work by Màrquez-Luna t al (pubmed.ncbi.nlm.nih.gov/29110330/) and Mostafavi et al (pubmed.ncbi.nlm.nih.gov/29110330/).

The bottom line: The increase in prediction accuracy is modest, but fairly robust. The approach does away with the awkward task of finding a "matching" cohort, which often does not exist for a given individual. It can also reveal unexpected gene-by-something interactions.

Each individual gets assigned a weighted mixture of scores based on age, ancestry, sex, and other information. This typically favours scores learned in groups of matched sex and ancestry, but not always. Age effects tend to be important, reflecting e.g. differences in medication use.

Personalized polygenic risk prediction and assessment with a Mixture-of-Experts framework With the increasing availability of high quality genomic data from diverse cohorts, polygenic scores (PRS) have become a mainstay of genetic analyses of complex traits and diseases. Despite their prol...

To predict genetic risk for an individual, which score should you use? The one learned from a cohort that best matches by ancestry? Or by sex, or age, or environment? Shadi Zabad proposes to combine scores taking all these into account. www.biorxiv.org/content/10.1...

Testing of a suspect's DNA against crime scene DNA is fine. Screening existing databases in search of relatives is debatable. Indiscriminate collection of new genetic data on individuals with no link to the crime scene is fucked up and should not be tolerated.

Ontario privacy commissioner urges guardrails on police gravitating toward ‘unbridled genetic surveillance’ Watchdog’s report warns new era of DNA-based criminal investigations calls for clearer rules and more transparency

The RCMP surreptitiously collected DNA from random individuals at a Kurdish festival based on the inferred Kurdish ancestry of a suspect, in search of a relative. www.theglobeandmail.com/canada/artic...
RCMP: "Police will use all tools available."
Let's make this tool unavailable.

Preprint out: A multi-ancestry genetic reference for the Quebec population.

2173 WGS and 29k genotyping arrays in a longitudinal biobank.

www.medrxiv.org/content/10.1...

The impact of ancestral, genetic, and environmental influences on germline de novo mutation rates and spectra - Nature Communications Here the authors analyze de novo mutations in >10,000 parent-offspring trios and find that ancestry and smoking independently associate with mutation rate, but that common genetic variants likely c...

Delighted to see this paper out, investigating factors affecting human germline mutation in a cohort of 10,000 families from the UK 100k Genomes Project.

www.nature.com/articles/s41...

Canada Excellence Research Chairs (CERC) - Montréal, Quebec (CA) job with University of Montreal (UdeM) | 12838461 Canada Excellence Research Chairs (CERC)

If you’re interested in moving your lab to beautiful Montreal Canada, this program offers chairholders funding of up to 1M$ a year for 8 years. See link for details.
www.nature.com/naturecareer...

My sister Élise will get her doctorate without a single peer reviewed paper! She'll be getting a doctorate, honoris causa, from @uqam.ca for her contributions to children's literature. Her books are often engaged, often silly, and always good reads. Very well-deserved ❤️🍉!

This is so outlandish that I had to go and read the original to see if there was any context I was missing. But no, you summed it up perfectly. I guess they singled those out as "things an antiracist would say", and therefore verboten?

Ma présentation a été annulée par l’Université de New York Mon «alma mater» a préféré jouer de surobéissance préventive pour ne pas froisser le gouvernement Trump.

Une présentation de la Dre Liu à NYU annulée pour faits dérangeants. Son commentaire plein de nuance ici
www.ledevoir.com/opinion/idee....

A conference by Dr Liu at NYU cancelled for including unpleasant facts. Her thoughtful commentary (in French):
www.ledevoir.com/opinion/idee....

Methods like this can help geneticists find unexpected patterns in their models of human history. However, it will also be tempting to over-interpret them (because they are so pretty and detailed!). Expecting heated arguments in the near future about how these should be interpreted!

A map of human migrations

It creates detailed illustrations of possible routes followed by our ancestors. We spent a lot of time looking at such maps from recent historical records in www.science.org/doi/10.1126/.... Impressive to see this done with from genetic data and going back thousands of years!

A geographic history of human genetic ancestry Describing the distribution of genetic variation across individuals is a fundamental goal of population genetics. We present a method that capitalizes on the rich genealogical information encoded in g...

Grundler et al propose a new method to infer the geographic position of all our genetic ancestors www.science.org/doi/10.1126/...
building on a method by Wohns et al www.science.org/doi/10.1126/....

My thoughts on the approach: www.science.org/doi/epdf/10.....

Using the ancestral recombination graph to study the history of rare variants in founder populations Gene genealogies represent the ancestry of a sample and are often encoded as ancestral recombination graphs (ARG). It has recently become possible to infer these gene genealogies from sequencing or ge...

Ancestral Recombination Graphs are an amazing tool to study genetic variation. In this preprint, Alejandro Mejia-Garcia shows that ARGs are surprisingly good at identifying founder variants, and help us pick apart their geographic distribution and phenotypic effect.

www.biorxiv.org/content/10.1...

Canada Research Chair (CRC) Tier II in Synthetic Biology and Biomanufacturing - Concordia University The department of Biology in the Faculty of Arts and Science at Concordia University invites applications for a tenure-track faculty position with the opportunity to apply for a Tier II Canada Researc...

A faculty position in Synthetic biology at Concordia University in Montreal, with possibility of funding from a Canada Research Chair: www.concordia.ca/artsci/about... .

Faculty position opening in bioinformatics at the University of Montreal! There is a great bioinfo and comp bio community in Montreal.

www.umontreal.ca/carrieres/pr...

Bizarrement ça me dit que je veins de Laval :) Je ne sais pas si d'autres Québécois sont aussi assignés à l'ouest de la France, ce qui serait cool.

Experience using conventional compared to ancestry-based population descriptors in clinical genomics laboratories While the use of genetically defined population descriptors in human research studies has been proposed to reduce inaccuracies and harms associated with the use of conventional, non-biologically based...

📢New Perspective from Hatchell et al.
📰Experience using conventional compared to ancestry-based population descriptors in clinical genomics laboratories
www.cell.com/ajhg/fulltex...

I posted a couple days ago about our new paper on building causal graphs from genetic associations + Perturb-seq.

Here I want to expand on the value of using DIRECTIONAL information contained in LoF burden tests.🧵
[work led by @minetoota.bsky.social ]

bsky.app/profile/jkpr...

Towards whole-genome inference of polygenic scores with fast and memory-efficient algorithms With improved Whole Genome Sequencing (WGS) and variant imputation techniques, modern Genome-wide Association Studies (GWASs) have enriched our understanding of the landscape of genetic associations f...

Extremely fast bayesian PRS calculation, by Shadi Zabad. Lots of clever tricks to improve memory efficiency, multithreading, and numerical robustness of WGS statgen analyses. www.biorxiv.org/content/10.1....

A beautiful news story on the Silent Genomes Project (in French) ici.radio-canada.ca/info/long-fo...

I'm getting a massive and sustained uptick in spam emails that are either

1-invitations to submit a book manuscript
2-offers to ghost write book manuscripts for me.

Do others see this, or did my email just reach an old spam list? Wondering if this is a new GPT-based ghostwriting model.

A leading grad school asked to rank a student by "emotional maturity", and options included "top 1%", "Top 3%", etc.

I already have all interns compete in the marshmallow delayed gratification experiment, but more than 3% succeed. Help!

Assistant Professor (Tenure Track), Department of Epidemiology, Biostatistics and Occupational Healt... Please refer to the How to Apply for a Job (for External Candidates) job aid for instructions on how to apply. If you are an active McGill employee (ie: currently in an active contract or position at ...

A TT position in biostatistics and medicine at McGill University:

mcgill.wd3.myworkdayjobs.com/en-US/mcgill...

En français: : mcgill.wd3.myworkdayjobs.com/fr-CA/mcgill...