Excited to present the first paper of my PhD: Robin, a multi-agent system for automating scientific discovery! 💡Robin was applied to identify a promising therapeutic candidate for dry AMD, ripasudil, which is a clinically-approved drug not previously used for dry AMD. 🧵👇

🚨 New preprint 🚨 We introduce Generative Distribution Embeddings (GDEs) — a framework for learning representations of distributions, not just datapoints. GDEs enable multiscale modeling and come with elegant statistical theory and some miraculous geometric results! 🧵

Want to improve your protein or genomic language model’s performance at zero-shot variant effect prediction? We propose a simple adjustment to likelihood-based approaches. biorxiv.org/content/10.110…

Can protein expression be ‘solved’? We review current methods for measuring and predicting protein expression (wet lab + ML techniques!) & propose how this ubiquitously interesting problem could be solved. What do you think? sciencedirect.com/science/articl…

Great blog post. All ML folks working in the bio space should seriously read this. This issue is seriously plaguing the field. Lots of bombastic AI4bio papers getting tons of attention (pun intended) that end up being entirely hollow under the surface. This is BAD for the field.

🚨ICML Paper Alert🚨 What if finding the right protein homologs wasn't a slow search, but a learned part of the model itself? We introduce 𝐏𝐫𝐨𝐭𝐫𝐢𝐞𝐯𝐞𝐫, an end-to-end framework that learns to retrieve the most useful homologs for self-supervised reconstruction! (1/12)

If you love #viruses, #ribosomes, and genomic #darkmatter, this thread is for you!! 💫 We're excited to share our published work developing Massively Parallel Ribosome Profiling (MPRP), which uncovered ~4,000 hidden proteins in ~700 viral genomes. science.org/doi/10.1126/sc…

Endometriosis is an incredibly interesting disease 5k words, 23 minutes reading time covering one of the strangest conditions ive ever heard about link: owlposting.com/p/endometriosi… very grateful to Shilpa Pothapragada for initial inspiration for this piece + reviewing it!!

There are many domain-specific noise processes for discrete diffusion, but masking dominates! Why? We show masking exploits a key property of discrete diffusion, which we use to unlock the potential of those structured processes and beat masking! Nate Gruver Andrew Gordon Wilson 1/7

🚨 New paper 🚨 RNA modeling just got its own Gym! 🏋️ Introducing RNAGym, large-scale benchmarks for RNA fitness and structure prediction. 🧵 1/9

I wonder if the gym membership for ProteinGym covers RNAGym as well?? oh wait, it's all free and 100% open source!

🚨 new paper alert! science.org/doi/10.1126/sc… During my PhD, one of the most frustrating challenges was trying to interpret genes labeled as “hypothetical proteins.” 1/n

We can make population genetics studies more powerful by building priors of variant effect size from features like binding. But we’ve been stuck on linear models! We introduce DeepWAS to learn deep priors on millions of variants! #ICML2025 Andres Potapczynski, Andrew Gordon Wilson 1/7

First CASP... now Machine Learning in Structural Biology (MLSB)... sad to see workshop getting cut by Neurips. 😭

Whose protein design model reins supreme? You design, we test, science wins 😛 Protein Engineering Tournament 2025 registration is live!!!!

Curious about why pretraining misses important features?🚩 Our new paper reveals a key limitation in how deep learning models learn – an information saturation bottleneck.🤯 Networks cannot learn new features after “saturating” information on related features. 💡(1/2)

Come see how to shrink protein sequences with diffusion at our talk tomorrow morning!!!

Congratulations to the entire @ProfluentAI team on this incredible milestone! OpenCRISPR-1 represents a paradigm shift - the first AI-designed CRISPR protein to successfully edit human DNA with fewer off-target effects. We're moving from discovery-based to engineered biology. 🧬

1/5 Biological data is noisy, redundant, and ever-growing. 🗣️ In our new paper (first paper of my post doc!! ⚡️), we track model performance across 14 years of UniRef100 snapshots to ask: how does pLM performance scale with training data?

Excited to share work with Zhidian Zhang, Milot Mirdita, Martin Steinegger, and Sergey Ovchinnikov biorxiv.org/content/10.110… TLDR: We introduce MSA Pairformer, a 111M parameter protein language model that challenges the scaling paradigm in self-supervised protein language modeling 🧵