Ryan Bailey

Excited to share our new preprint: ‘Divalent HIV gp120 Immunogen Exhibits Selective Avidity for Broadly Neutralizing Antibody VRC01 Precursors’! We’ve designed a vaccine immunogen that binds divalently to target B cell receptors (like VRC01) but only monovalently to non-target BCRs. Check it out: biorxiv.org/content/10.110… 1/7

I'm still not super worried about a hantavirus pandemic, but there's also no technical reason this shouldn't be the standard timeline for new threats: ~Hours: definitive confirmation via sequencing ~Days: candidate mRNA vaccine designs ~Weeks: millions of vaccine doses

What happens when you let frontier LLMs design proteins, and then synthesize and test them in a wet lab? We ran a protein design competition with @muni_bio where AI agents competed against humans to design molecules that bind TREM2, a key receptor linked to Alzheimer’s. Results: GPT 5.2 and Grok 4.1 both placed in the top 5, with molecules showing strong binding to TREM2 when tested in our lab.

This week's episode of viroLOGICAL is about Andes virus. (sorry for my voice, I was battling a respiratory infection) ➡️ Spotify: open.spotify.com/show/2LxUpjumK… ➡️ Apple Podcasts: podcasts.apple.com/at/podcast/vir… ➡️ Podigee: virological.podigee.io

Super clear explanation of Life Biosciences’ approach to epigenetic rejuvenation

RevMed is such a cool story. The trip from academia to a failed biotech (WarpDrive) to huge success is a microcosm of the circuitous path often taken in drug development. nytimes.com/2026/05/12/hea… via @NYTimes

Frontier LLM agents can beat a hand-engineered protein design pipeline. So why do they still lose to a human expert? New preprint from @romerolab1 — BioDesignBench: 76 expert-graded protein design tasks. DeepSeek V3 & GPT-5 beat it. The expert still wins. Why? (1/4)

@sama Why pregnant pause now before responding even with instant?

what would you most like to see improve in our next model?

Introducing Genie 3, a generative protein model that substantially advances the state-of-the-art for binder design, increasing in silico success rates by up to 20x on hard multimeric targets. It also debuts a form of inference-time scaling unobserved in other design models. 🧵1/8

Notes on my trip to Shanghai. On how China has temufied all of biotech. Investigator-initiated trials. And why we are choosing to run trials there. Link in comms

Your brain doesn't form the thought until you write it down. Nature Reviews Bioengineering published the case for that claim last summer in an editorial titled "Writing is thinking." The cited evidence is a 2024 EEG study at the Norwegian University of Science and Technology. 36 students alternated between handwriting and typing the same words. 256-channel sensor array. Cursive on a touchscreen versus keys on a keyboard. Same words both ways. Handwriting produced widespread connectivity across parietal and central brain regions. Typing didn't. The theta and alpha frequency bands the literature ties to memory formation and encoding lit up almost exclusively when the hand was forming the letters. The motor act was producing the cognition. What the editorial extends from that finding is the more uncomfortable claim. Writing a scientific article is the mechanism by which a researcher discovers what their main message actually is. The act of constructing sentences forces the chaotic, non-linear way the mind wanders into a structured, intentional narrative. You sort years of research into a story, and in the sorting, you find out what you believe. Then the line: If writing is thinking, are we not then reading the thoughts of the LLM rather than those of the researchers behind the paper? Nature endorses LLMs for grammar, search, brainstorming, breaking through writer's block. Where the line gets drawn is outsourcing the whole writing process. Because the writing process is the thinking process. Even editing the LLM's draft is harder than writing one from scratch. To restructure someone else's reasoning you have to reconstruct it first, which means doing the cognitive work anyway, with worse leverage and more friction. The time savings on the keyboard turn out to be cognitive savings on the part of the brain you wanted to use. Your first draft was the thinking.

How it feels like to get ratioed for a bad take

mRNA delivery of mosaic-8 pan-sarbecovirus RBD vaccines elicits distinct antibody epitope signatures @CellReports @bjorkmanlab cell.com/cell-reports/f…

New online! Hypervariable loop profiling decodes sequence determinants of antibody stability dlvr.it/TSJN2Y

Can we use #proteindesign to develop countermeasures before viruses cause pandemics? We computationally designed an ultrapotent and broadly neutralizing deltacoronavirus miniprotein. Led by Nathan Avery @hhmi_science 1/12 pnas.org/doi/10.1073/pn…

AI-Driven De Novo Binder Design: From Structure Prediction to Closed-Loop Optimization spj.science.org/doi/10.34133/c…

The past, present and future of denovo protein design nature.com/articles/s4158…

Patrick Winston taught engineering at MIT for 50 years. This is his last lecture. He died 5 months after recording it. It was his final gift to the world. 1000s of the greatest minds passed through his classroom and went on to engineer the world. He has some wisdom…

@davidasinclair @elonmusk That’s why X is so great. Hear it straight from the source.

Anyone who knows me knows that I’m very careful with my words and my research What undermines public trust isn’t scientists, it’s the selective editing of clips and the constant hit-job reporting by mainstream media I understand why @ElonMusk stopped engaging

AI can now design antibodies that bind with atomic precision, but not ones that cells can produce. Our preprint closes this gap, delivering a structural principle, an AI-guided rescue pipeline, and adalimumab variants with 20-100x in vivo potency. biorxiv.org/content/10.648…

Discovery of a new receptor by which alphacoronavirus from bats can get into human cells and the potential for transmission nature.com/articles/s4158… @Nature nature.com/articles/d4158… @NatureNV