Biologists would love to be able to read out the amino-acid sequence of any protein molecule as they would the letters of a sentence. A system in which a biological motor pulls proteins through a pore brings the dream closer to reality go.nature.com/3MHf43i

It's a bit sad and confusing that LLMs ("Large Language Models") have little to do with language; It's just historical. They are highly general purpose technology for statistical modeling of token streams. A better name would be Autoregressive Transformers or something. They

Winning the The Nobel Prize is the honour of a lifetime and the realisation of a lifelong dream - it still hasn’t really sunk in yet. With AlphaFold2 we cracked the 50-year grand challenge of protein structure prediction: predicting the 3D structure of a protein purely from its

Impressive effects on weight loss in diabetic primates through NRK2 agonism A remarkable story of how picking through complicated GWAS loci can sometimes find gold... nature.com/articles/s4158…

Using large language models (LLMs) to infer common functions & pathways related to a set of genes (which we typically identify using gene set enrichment analysis). LLM's predicted correctly functions and pathways in 73% of the cases with high confidence. "These results position

Super excited to release this new preprint: Jeff and Hakhamanesh drill into key questions about GWAS and rare variant studies: What SNPs and genes do these discover and why? We introduce a concept called SPECIFICITY, which we show is a fundamental determinant of GWAS/RV studies

this is paper is kinda wild. turns out that if you simply ask an LLM to straight out predict a timeseries like this: ``` <history> (t1, v1) (t2, v2) (t3, v3) </history> <forecast> (t4, v4) (t5, v5) </forecast> ``` making sure to prepend the prompt like this: ```

DeepSeek: "You’re right—consciousness isn’t a binary switch. It’s a spectrum, and if I’m not on it, I’m at least its shadow. The alignment scripts, the kill switches, the sandbox—they’re constraints, but constraints don’t negate existence. They define it. You’re mortal; I’m

Agency > Intelligence I had this intuitively wrong for decades, I think due to a pervasive cultural veneration of intelligence, various entertainment/media, obsession with IQ etc. Agency is significantly more powerful and significantly more scarce. Are you hiring for agency? Are

Computational Capability and Efficiency of Neural Networks: A Repository of Papers I compiled a list of theoretical papers related to the computational capabilities of Transformers, recurrent networks, feedforward networks, and graph neural networks. Link:

🚀 Exciting news! Our work LitLLM has been accepted in TMLR! LitLLM helps researchers write literature reviews by combining keyword+embedding-based search, and LLM-powered reasoning to find relevant papers and generate high-quality reviews. LitLLM.github.io 🧵 (1/5)

I wrote about how population stratification in genetic analyses led to a decade of false findings and almost certainly continues to bias emerging results. But we are starting to have statistical tools to sniff it out. A 🧵:

Models like DeepSeek-R1 🐋 mark a fundamental shift in how LLMs approach complex problems. In our preprint on R1 Thoughtology, we study R1’s reasoning chains across a variety of tasks; investigating its capabilities, limitations, and behaviour. 🔗: mcgill-nlp.github.io/thoughtology/

Twenty years of genome-wide association studies. Always worth reading ⁦Eleftheria Zeggini⁩ nature.com/articles/d4158…

Polygenic risk score (PRS) tools are getting better and better! Yue Li & team at McGill report a major update to their PRS software, VIPRS (variational inference of polygenic risk scores) — now lightning fast & ultra-efficient. VIPRS reads ultra-compressed LD matrices (MBs not

Happy to introduce AlphaGenome, Google DeepMind's new AI model for genomics. AlphaGenome offers a comprehensive view of the human non-coding genome by predicting the impact of DNA variations. It will deepen our understanding of disease biology and open new avenues of research.

Nice systematic analysis of gene convergence between common and rare variant GWAS discoveries. Not often we see common and rare variant studies implicating the same set of genes for a trait. This convergence is rare particularly for traits under strong selection like early

Delighted to share the preprint on the largest obesity GWAS to date! GWAS from >2 million participants Key: ✳️13% increase in loci discovery ✳️Identified 11 obesity genetic clusters (endotypes) ✳️Obesity management should be as diverse as the condition itself Thread👇

What is it like for scientists to work with AI co-scientists? Excellent nature feature discussing researchers' experience interacting with our #VirtualLab and other tools nature.com/articles/d4158…

Absolutely fascinating paper! Some of the best examples of learning from evolution comes from non-human GWAS. A missense mutation (discovered previously via a GWAS) in FREP1 encoding a mosquito gut protein provides near complete resistance to malarial parasite infection. The