🚀 Introducing Nucleotide Transformer v3 (NTv3) Today, we are very excited to share our latest foundation model for biology - Nucleotide Transformer v3 (NTv3). NTv3 is InstaDeep new multi-species genomics foundation model, designed for 1 Mb, single-nucleotide-resolution

Mark your calendars: The AI x Bio meeting of 2026 will be held at CSHL on May 26-31! The program brings together 50+ invited leaders in genomics, transcriptomics, protein design, drug discovery, neuroAI, pathology, agentic AI, and more! Abstract: 3/26 meetings.cshl.edu/meetings.aspx?…

Interested in pursuing a PhD at the intersection of AI & genomics? The Koo Lab is recruiting through the new BioAI PhD Program at CSHL! Send me an email w/ CV. Applications are rolling until position is filled. *A Master’s degree or equiv is required cshl.edu/phd-program/bi…

With all the buzz around Claude Code, I thought about switching from ChatGPT, but then I hit cancel and got offered a free month. One month to see if it can keep up!

x.com/i/article/2013…

Okay so, we just found that over 50 papers published at @Neurips 2025 have AI hallucinations I don't think people realize how bad the slop is right now It's not just that researchers from Google DeepMind, Meta, Massachusetts Institute of Technology (MIT), Cambridge University are using AI - they allowed LLMs to generate

AlphaGenome is out in @nature today along with model weights! 🧬 📄 Paper: nature.com/articles/s4158… 💻 Weights: github.com/google-deepmin… Getting here wasn’t a straight path. We sat down @googledeepmind to discuss the story behind the model, paper & API: youtu.be/V8lhUqKqzUc

Toward Interpretable and Generalizable AI in Regulatory Genomics 1 The review reframes seq2func models as living systems that improve through continual AI–experiment feedback loops instead of one-off training, arguing that static benchmarks hide systematic failure modes. 2 It

Excited to share our new Review/Perspective on interpretable and generalizable AI for regulatory genomics. Grateful to be a co-first author with an amazing team. Hope it’s useful to the community!

Excellent forward looking perspective from Peter Koo . While this is not stated in the perspective & I don't want to put words in Peter's mouth, I think this is an alternative vision to building "virtual cells" that aims to unify cis & trans regulation. 1/

1/13 Excited to share our (anna spiro Maria Chikina Sara Mostafavi) latest preprint! 🧬💻 Personal Genome Prediction isn't just a downstream task—it’s the ultimate end-to-end benchmark for Variant Effect Prediction. We put the new SOTA AlphaGenome to the test and uncovered a

My bio says I work on AGI preparedness, so I want to clarify: We are not prepared. Over the last year, dangerous capability evaluations have moved into a state where it's difficult to find any Q&A benchmark that models don't saturate. Work has had to shift toward measures that

Nice seq2func interpretability approach that assigns attribution to each output bin and visualizes the flow of influence from sequence to outputs, avoiding the loss of info from collapsing the profiles into a single value!

[Postdoc Opportunity] Interested in research at the intersection of AI, regulatory genomics, and plant biology? Ware Lab and Koo Lab at CSHL are seeking a highly motivated postdoc fellow to pioneer this space! - Fully funded: Must be US citizens! - DM or email me CV Please RT!

Check out this latest proof-of-concept regulatory DNA LM (called ARSENAL #GGMU) by Aman Patel that is pretty much the opposite of the current trends in DNALM literature. (Beware: Long thread) 1/

excited by this result. Tang et al (koo lab) had an amazing work evaluating models dna-lm embeddings where they found little to no benefit in using dna-lm embeddings across multiple tasks. This seemed to have worked.

Are neural nets across modalities really converging to the same representation as they scale, as the Platonic Representation Hypothesis suggests? We show that common representational similarity metrics are confounded by network width & depth. We propose a permutation-based