Late-interaction retrieval is incredibly powerful, but scaling it is computationally challenging. k-means is a huge bottleneck. Our new architecture, TACHIOM, is fully open-source and tackles this problem: up to 247x faster clustering and 9.8x faster retrieval. ⚡ 🧵👇

Beware #RStats users, according to Misanthropic, an R to C transpiler is a security risk 😄 Fable turned out, as expected, to be La Fable I think some french speaker at Misanthropic is doing some kind of joke with these model names

Arguably the most boring step in genomics is the first one: normalization. Settled science. Scale log. Move on. Except that here's been a huge blind spot in the field. And it matters for AIxBio. A 🧵about what I think may be one of the most important papers I've written. 1/

Find it incredible that REVEL remains undefeated

One billion 3D protein structures: Should you care about ESMfold2? youtube.com/shorts/ucRZwlQZc…

"What if I took a physics test in French and a French test in physics?" Apparently the French language is not compatible with physics. aifails.substack.com/p/physi…

I think the basic issue in all such discussions is that these kinds of competencies aren't some kind of essential "human attributes". Measuring cognitive properties by humans is myopic and causes all kinds of pseudoproblems. The emerging field of diverse intelligence has better frameworks. A spectrum of highly variable intelligences, and lots of research on which kinds of architectures enable which kinds of patterns (of behavior, of computation, of physiology, etc.), is more useful for discussions of natural, artificial, and hybrid agents of varying provenance and composition. frontiersin.org/articles/10.… journal.frontiersin.org/arti… and more at drmichaellevin.org/publicati…, plus lots of other good labs.

Quite nice crate ! so i made un bebe vibeslop R binding with some generic hacks i use for multiversion SIMD support in external C bindings (where cargo-multivers does not apply github.com/ronnychevalier/ca…) github.com/sounkou-bioinfo/R…

notes from my cascadiajs talk cascadia.wzrrd.sh/

To all folks dealing with biological data: do you ever need to check if your reads contain barcodes/adapters/primers/...? Or off-target matches? Sassy is the tool to use! A super fast implementation of "approximate string matching" with a grep-like CLI. curiouscoding.nl/posts/sassy…

Good software is invisible. That's why all software is bad because the managers demand visibility of programmers work.

been working through lots of quadratic equations lately. not in school or anything, just exploring free will :) made a video about the quadratic formula, perfect squares, and DIY implementation in R. #rstats you might find it interesting, idk. youtu.be/CRyBiJfh9sU?si=_TdN…

Our study led by @Yunfeng_Ruan describes a method of interpolated polygenic risk scoring (DiscoDivas) for more accurate scoring across ancestries, particularly in settings of admixture sciencedirect.com/science/ar… @AJHGNews @AniruddhPatelMD @skoyamamd @buutrg @somijemmacho @HornsbyWhitney @AndrewHaoyu @nilanjan10c

Can CPUs outperform top-tier GPUs in protein folding? 🧬 Thrilled to share our new bioRxiv preprint: "Revisiting CPUs for Protein Folding: Xeon-Based Acceleration of AlphaFold2." We achieve a 1.76x speedup for end-to-end pipeline over an H100 GPU baseline (FastFold)! 🧵👇 (1/n)

I'm super excited to share a tool I've been working on for a while now-- a full stack population genetics library that is purpose built for GPUs -- pg_gpu biorxiv.org/content/10.64898…

Pounce - a pure-Rust, Ipopt-inspired interior-point nonlinear programming solver (with pure Rust linear algebra solver) is in the wild! github.com/jkitchin/pounce #rust #optimization

🔗link to the paper: nature.com/articles/s41586-0… 📊summary statistics: GWAS catalogue ebi.ac.uk/gwas/ accession numbers GCST90449363–GCST90451603

Analyzing phenotypes on the wrong scale reduces power and creates spurious genetic interactions. SIQReg, a new method from my postdoc Zhenhong Huang, learns an optimal phenotype scale to fix this: biorxiv.org/content/10.64898…

DistPCA: Tera-Scale Genomic PCA via Out-of-Core Distributed Parallelism biorxiv.org/content/10.64898…

More fun :D