Alexey Sergushichev

I'm happy to present our new web-tool CORESH alserglab.wustl.edu/coresh/ for searching public gene expression datasets using gene signature as a query, developed by @vd_sukhov, which was just published in @NAR_open doi.org/10.1093/nar/gk… 1/n

Not sure how to feel about it, but someone rewrote—I guess with Claude Code or something—fgsea to Rust with Python bindings. Not an endorsement, but I know some people were interested in using it in Python, so here it is: github.com/deminden/rsfgs…

@strnr I find it funny that it still looks cheaper to buy the book

I normally bristle when I hear dismissive oversimplifications like "plagiarism machines" and the like. But things really are (expensive!) plagiarism machines in a sense. Extracting books from production language models arxiv.org/abs/2601.02671

@vsbuffalo I wonder when will we get a virus in the form of LLM prompts that's going to hijack people's API codes (or even pay for itself) to copy itself to new machines.

Given AI advances in hacking, is it time to start using Yubikeys for authentication on everything? I can’t tell if I’m overly anxious, or this is just realism.

@slavov_n I mean that achieving IMO gold sounds pretty realistic. And I expect it's going to be tested this year in a proper environment, like it was done for ICPC.

@assaron Computing systems have a long history of exceeding human performance in specific tasks, even before the time of the pocket calculator. LLMs have excels in new areas. Yet, this does not justify the claim of reaching AGI or the misleadingly incomplete report on IMO problems.

A reminder that the Turing test is just a simple suggestion.

@slavov_n LLMs definitely achieved better then human performance in competitive programming x.com/relizarov/stat…

The Nature comment asserts that LLMs have achieved gold-medal performance at the IMO. This is a contested claim, not a fact. - This is a self-reported claim confounded by allegations of data leakage. Let's talk about High School Math: arxiv.org/abs/2510.26768

Just noticed the "Samuel LeCun Jackson", what a cross-over!

@lisyarus @shipilev I got this problem frequently with my browser, but then learned that Firefox has an option to not close window with the last tab. Now I can press and hold Ctrl-W to my satisfaction!

@shipilev I regularly use holding Ctrl+W to close many opened tabs, but doing this with no tabs open closes the app itself, and ofc it's very easy to miss that last closed tab and accidentally shut the whole thing down.

VSCode "go to last edit location" is ^K^Q, and "quit VSCode" is just ^Q. Brilliant default key mappings right there. /me reopens VSCode for third time in the row...

#aging #rejuvenation #outlook #Community @NatureAging just published a very nice piece describing the view on the field of aging from ~30 experts in the field, humbly yours including: nature.com/articles/s4358… In addition to reading the key messages that were nicely assembled by Sebastien Thuault and Hannah Walters, i strongly recommend to read supplementary material to that article, where each of the authors more extensively asnwers to these 10 questions - it is incredible source for discussions, debates and raw insights: 1. Is there one advance in aging or age-related disease research from the past 5 to 10 years that changed how you think about the field, and why? 2. What have we learned about translating geroscience from model organisms to humans, and where do the biggest gaps remain? 3. The field of aging is very broad, covering biology, clinical, public health and social sciences. Has your work or thinking been inspired by approaches or findings from separate disciplines? 4. Which single shared resource (e.g. dataset, biobank, model, tool) would most accelerate progress in your field? 5. How should we balance large, collaborative team science efforts and the focus and agility of individual labs to drive research forward? Should there be more big-team science in aging research? 6. Where do you see research on aging and age-related diseases having the biggest impact on clinical care and public health now and in the future? 7. If you could change one funding or regulatory policy to speed up progress, what would it be and why? 8. There is growing public interest in aging and age-related disease research. What can researchers do to ensure that aging science can be trusted and benefits everyone? 9. What advice would you give to researchers entering the field now? 10. Where do you see your field heading in the next 5 to 10 years?

More info and examples are on the package website: alserglab.github.io/mascarade/ 3/3

For this release I optimized the performance and simplified working with Seurat objects 2/3

mascarade package for generating borders around clusters on UMAP/t-SNE plots is available on CRAN now! 1/3

Very excited to announce our latest paper in Nucleic Acids Research: "Accurate chromatin marks peak calling with Omnipeak." NAR Nucleic Acids Research lead by Oleg Shpynov lnkd.in/gqTJ4k5D For years, we all have been forced to swap between different tools depending on their data: MACS2 for sharp peaks or SICER for broad domains. Furthermore, the same methods could give vastly different numbers of peaks for the same mark when the quality varies a bit. Omnipeak changes the game: ✅Unlike MACS2, which struggles with broad histone modifications, Omnipeak captures the full architecture of the epigenome—from razor-sharp H3K4me3 peaks to expansive H3K27me3 domains—within a single framework. ✅ Better Biological Signal: Our method demonstrates higher consistency across biological replicates, ensuring that the "peaks" you find are real biological features, not technical noise. ✅ Superior Boundary Precision: In our benchmarking against ENCODE and Roadmap datasets, Omnipeak consistently outperformed existing methods in defining exact peak start/end coordinates. ✅ Consistency at a low-quality peak calling and no dependency on the control samples. This also provides unique systematic approach for any large-scale epigenetic analysis of many studies and samples. Stop juggling multiple pipelines. Switch to a universal, unsupervised solution that brings mathematical rigor to your ChIP-seq and ATAC-seq analysis.

@shipilev Leet!

That feeling when you had the regular update script running at one of your public facing servers, but you also forgot to reboot it for 3.5 years. I am sure nothing bad is going to happen if I do it now, 2 days before NY break starts.

1/ New in @NatureComms: A scalable method for non-additive GWAS reduces computational cost 700-fold and uncovers 781 novel disease loci in @FinnGen_FI . Led by @ivan_molotkov_ , this work opens large-scale dominant/recessive analysis for large biobanks. nature.com/articles/s4146…

🚀 New preprint! The Human Cytokine Dictionary - 9.7M cells, 12 donors, 90 cytokines - our largest human single-cell perturbation atlas yet. Users can map cytokine activity in their own data. Huge thanks to Lukas, Sören, Larsen, Parse Bio & Seelig lab! 🔗 biorxiv.org/content/10.648…

@ChShersh You can try to post a corrected one next year and compare the numbers of likes!

@assaron This explains everything

My most underrated CS joke

Bioinformatics is in shambles at the Denver airport’s acronym game

Very happy to see cover on Nov 11 @ImmunityCP issue highlighting our review on Human Immune Aging (sciencedirect.com/science/articl…). One of key visuals highlights major finding made by Terekhova et al in Immunity, 2023 - systematic shift toward type 2 (Th2/Tc2) immunity in healthy aging! (sciencedirect.com/science/articl…)

Just out - our major review piece @ImmunityCP , summarizing couple of decades of the research on human immune aging and providing highlights of the latest advances in the field. Led by Marina Terekhova, truly encyclopedic depth (327 references) sciencedirect.com/science/articl…