Decoding immunity. Autoimmunity is far more common than anyone today expects or understands. CSO @cdi_labs
Oregon
Joined January 2018
- Tweets2.7K
- Following1.2K
- Followers1.4K
- Likes7.8K
Tyler Hulett retweeted
General prediction of T cell receptor antigen specificity from sequence using AlphaFold 3
science.org/doi/10.1126/scia…
Tyler Hulett retweeted
My daughter has an ultra-rare, autosomal dominant, monogenic neurological disorder with fewer than 200 cases worldwide. It is characterized by haploinsufficiency; however, unlike some genes underlying other genetic diseases, the affected gene is too large to fit into a standard AAV vector for a gene therapy intended to deliver neurons a working copy. I am using GPT-6 Astra and the Rosalind Workbench to design candidate minigenes for my daughter (and others affected) that are sufficiently sized for viral delivery while retaining the most important functional domains of the protein. The idea of minigenes has precedent: Elevidys by Sarepta Therapeutics is FDA approved for Duchenne Muscular Dystrophy and consists of a “micro-dystrophin” transgene about a third of the size of the wild type dystrophin. Specifically within Rosalind I am using the Biological Sequence & Alignment Viewer, Biohub ESM plugin (for folding candidate minigene variants and domains), and the Molecular Structure Viewer (for alignment to the wild-type reference predicted structure). Furthermore, I am leveraging additional databases like ClinVar to cross-reference pathogenic missense variants that suggest regions critical to functional activity, since little is known about the disease in general. I would be exceptionally thankful for the additional credits of the Pro plan as I keep quickly running out of usage, and this is a use case where I most certainly want the best available model.
ALT Minigene variants designed by GPT-6 Astra
Tyler Hulett retweeted
This is legitimately terrible for OpenAI --- but also Anthropic --- and it relates to their work in drug discovery.
I read this as: prompts will be monitored and potentially acted on if we think there’s something scientifically and/or commercially valuable being disclosed.
Now, the big problem for Anthropic is that they are increasingly seen to be competing with their customers. Outside of SF, many people also think of OpenAI/ChatGPT and Anthropic/Claude as somewhat interchangeable companies and products --- aka, if OpenAI do something, surely Anthropic will do it too?
As so much IP relates to drug targets and people disclose these to Claude, I think we will see a lot of pharma companies stop using these tools or put up a lot of internal guardrails.
wtf is this way to handle mathematicians work and scientific communication
TLDR: Leven and Tristan worked over several months on one of the Millenium Prize Problems with various AIs to reach final interesting results. OpenAI apparently heard about it in the last days and prompted their latest models to work on the direction Leven and Tristan found fruitful. They then tried to push for controlling communication of the result and dropping Leven from authorship with some very bad taste social pressure.
Hope this is not a glimpse of the future we’ll get in science research with these dominating players playing marketing games hurtful for the real scientific community.
Tyler Hulett retweeted
seems like the general timeline to get medicine to stop making an active mistake* is ~25 years
* eg, autologous bone marrow transplants after total body irradiation for cancer treatment (oops cancer cells hiding in there), anti-menopause hormones, "have you tried losing weight?"
Doctors X-rayed pregnant women routinely in the 1950s to check which way round the baby was lying. Nobody thought it was dangerous.
Alice Stewart noticed childhood leukaemia rising sharply in Britain and set out to find why. She and her team interviewed the mothers of 1,400 children who had died of cancer, and 1,400 mothers whose children were alive, and asked them the same questions.
The pattern showed up in the first 35 pairs. Children who had died were about twice as likely to have been X-rayed in the womb, and a single scan at the dose used then appeared to be enough.
She published in 1956, and the response was outrage. Radiation was the technology of the future in the 1950s, and the claim that a low dose could cause cancer years later contradicted what the entire medical and nuclear establishment believed. Her methods were attacked. Her funding suffered. Doctors carried on X-raying pregnant women.
It took roughly 25 years for the medical world to come round. Pregnant women were still routinely X-rayed into the late 1970s. Two decades after she told them to stop. She retired in 1974 and did her most contested work afterwards, in her seventies, examining the health records of American nuclear weapons workers. Her findings there were unwelcome enough that the Department of Energy cut off her access to the data. She spent 14 years fighting them and testified before Congress, and in 1990 the records were opened to independent researchers.
She died in 2002, aged 95. The children who were never X-rayed have no idea who she was.
history.rcp.ac.uk/blog/alice…
Tyler Hulett retweeted
What if the proteome encodes more biology than we can see from proteins alone?
What if protein degradation is not simply cellular disposal, but also a source of biological function?
Every cell is continuously breaking down and remodeling proteins, generating a vast and dynamic pool of peptide fragments. We propose that some of these fragments are not inert byproducts, but bioactive molecules with important roles in biology.
In our new Essay in @PLOSBiology, “Encrypted immunity mediated by peptides,” Zhenrun J. Zhang, @YMerbl, and I explore this idea through the emerging framework of encrypted immunity: bioactive peptides hidden within larger proteins that can be released through proteolysis and contribute to antimicrobial defense, immune signaling, inflammation, and potentially broader aspects of physiology.
AI and machine learning are helping us explore this hidden molecular space at a scale that was previously inaccessible, uncovering candidate peptides in humans, microbes, archaea, venoms, and even extinct organisms.
But identifying sequences is only the beginning.
The deeper questions are biological: Which peptides are actually produced in vivo? When and where are they generated? How is their release regulated? What functions do they perform? And how do we distinguish meaningful biological signals from the background of protein turnover?
Answering these questions will require computation and experimentation to advance together.
To me, this points to a broader opportunity for AI in biology: using machines not simply to predict biology, but to help us uncover its hidden organizing principles—and ultimately learn how to engineer them for human health.
If this framework holds, it could deepen our understanding of how biological information is encoded and deployed, while opening new directions for therapeutics and diagnostics.
Much remains to be learned, and I’m excited about what comes next.
Weizmann Institute of Science, @Penn, @CBE_Penn, @PennBioeng, @PennEngineers, @PennMedicine, @PennChemistry, @PennMicro, @PennPsych, @PennSAS.
journals.plos.org/plosbiolog…
Tyler Hulett retweeted
Anti-CD320 Autoantibodies and Central Nervous System Vitamin B12 Deficiency in Idiopathic Myelopathy
Congrats to @jvpluv , Michael Wilson, Carlos Pardo and colleagues!
jamanetwork.com/journals/jam…
Tyler Hulett retweeted
Honored to share our Nature paper. In >1,000 hospitalized COVID-19 patients, ~50% had reactivation of dormant viruses creating a "dysvirosis" linked to worse acute COVID-19 and Long COVID outcomes. nature.com/articles/s41586-0… #LongCOVID #DellMed
Tyler Hulett retweeted
BoltzGen preprint was updated to include proteome-wide screening for off target interactions for one design (against IDI2) using CDI HuProt. The results were crazy clean - no off target binding (>21,000 targets)! With n=1 it's just a demo but I hope this becomes more common
Tyler Hulett retweeted
1/ We are thrilled to share our first @CellCellPress paper with the ATLAS team studying the role of antibodies, emphasizing autoantibodies in cancer (authors.elsevier.com/a/1nUSA…).
Tyler Hulett retweeted
🔆🔆🔆Matching T-cell receptors (TCRs) to their cognate peptides presented on MHC molecules (pMHCs) from sequence data alone is one of the most important unsolved problems in computational immunology. I've been presenting this work for over a year and am excited to finally share the preprint and public code.
A transformative collaboration for our lab with @quaidmorris (see tweetorial 👇👇with links). Led by first authors @Olga_Lyu, Jonathan Levine & Melissa Pathil, with S. Martis, A. Streltsov, Z. Sethna, Y. Elhanati (@elhanati_yuval), Vinod Balachandran (@TheVinodLab)!!!
@MSKCancerCenter
Can we predict which T cells recognize which peptides? A lot of the signal is hiding in AlphaFold3's hallucinations.
enFoldX: instead of one predicted structure, we use the whole noisy ensemble to predict TCR:pMHC binding.
💻github.com/jonlevi/enFoldX
biorxiv.org/content/10.64898…
Tyler Hulett retweeted
Correct.
Viruses are scary invisible, replicating killing machines.
Nature is exploring the space of possible human viruses every day by the many, many trillions of designs looking for better versions via evolution. Bad actors can use such viruses as weapons of terror.
We are undefended from this because we don’t have good biosecurity tech (we should detect viruses like we detect smoke and put the fire out fast) not because of a lack of guardrails on AI models.
Making AI models not work for doing biotech is a distraction from what we should be doing in biosecurity. In fact, it hurts as it slows the development of our countermeasures by blocking our researchers from the best AI tools.
Bio-weapons are hard. Even with AI, the difficulty of building one is quite substantial. Testing it is even harder. And the design space is full of trade-offs that make it hard to create a pandemic (let alone without human trials). Solid thinking as usual from @Abi0lvera
Tyler Hulett retweeted
THIS IS THE BEST PAPER OF 2026 SO FAR, A MONUMENTAL BREAKTHROUGH, A REVOLUTION IN THE FIELD OF INFLAMMATORY CONDITIONS:
Interleukin-10 Autoantibodies and HLA-DRB1*01:03 in Inflammatory Bowel Disease | New England Journal of Medicine nejm.org/doi/10.1056/NEJMoa2…
Tyler Hulett retweeted
Thrilled to share our new preprint: "AEGIS reveals epitope- and clone-resolved convergence of CNS B and T cell autoreactivity in ROHHAD." Important implications for mechanistically understanding autoimmunity! Massive team effort (attributions at end).
biorxiv.org/content/10.64898…
Tyler Hulett retweeted
1/ We are thrilled to share our collaborative study of B cells producing auto-Abs neutralizing type I IFNs in @CellPressNews (cell.com/cell/abstract/S0092…).
Tyler Hulett retweeted
Just finished a watershed board meeting at Centivax.
CEO Jake Glanville is holding the first vial of universal vaccine, now in clinical trials. In the long arc of human history, pathogens have killed >50% of humans that ever lived — over 50 Billion dead. Nothing else comes close.
This huge painting by Andrew Turner adorns the entryway. We have better tools than ever before to eradicate pathogens. We are hopeful that Centivax's universal vaccine programs will end the scourge of malaria, herpesviruses, coronaviruses, influenza and more.
Remember the variants? And vaccines that keep changing and don’t work so well? It’s not just COVID and coronaviruses, but flu, malaria, HIV and herpes/shingles too.
This happens because the pathogen is constantly mutating most of its surface coat proteins to evade our immune system — it keeps looking different. But there is an invariant element to each, and if we could only guide our immune system to see that signal in the noise, we’d have a “universal vaccine” for all variants, both known and unknown. Imagine a single flu shot that worked well every season and would also work for new variants, even scary ones like H5N1 bird flu, weaponized flu or the Spanish Flu were it to reemerge. It could end pandemics. It might even eradicate certain pathogens altogether, as we did with smallpox.
This has been a holy grail in vaccine development, one that I have philanthropically supported for many years (but those engineered nanoparticle approaches failed). Meanwhile, Centivax may have figured it out. Their approach has worked beautifully in many animal species, and human trials have just begun. We will know soon because there is a quick HAI assay that can evaluate the vaccine’s breadth of efficacy.
"For decades, flu vaccination has been reactive," said Sawsan Youssef, PhD, founder and Chief Science Officer of Centivax. "A universal influenza vaccine allows us to be proactive—moving from annual guesswork to predictable durable response."
“Beyond its flagship universal flu program, Centivax's epitope-focusing platform is advancing a growing pipeline spanning a pan-herpes Alzheimer's preventative, a broad oncology treatment, a malaria vaccine, and a universal antivenom recently published in Cell” — News today: prnewswire.com/news-releases…
Yeah, one universal antivenom shot for all snake species. It should also work for a variety of parasites: viral, bacterial, protozoan, even fungal outbreaks for the Last of Us. One shot to end each of them.
And an Alzheimer’s preventative? If we can avoid infection by herpesvirus and flu, large natural experiments suggest that this would be neuroprotective for Alzheimer's and Parkinson's. It may prove to be the most effective treatment for dementia and neurodegeneration. See nitter.cf/FutureJurvetson/status…
Fingers crossed that this works in the current flu trials, and then, applying it more broadly, Centivax may end the pandemic era.
Tyler Hulett retweeted
Viral infection and brain inflammation with seizures in PARK7 deficiency rupress.org/jhi/article/2/2/…
Tyler Hulett retweeted
There's an entire parallel scientific corpus most western researches never see.
Today i'm launching chinarxiv.org, a fully automated translation pipeline of all Chinese preprints, including the figures, to make that available.
Tyler Hulett retweeted
Exciting paper. Of 1140 Dutch patients with antibody-associated neurological disease, 36 had catatonia.
Of these 36, 35 had NMDAR encephalitis and 1 had KLHL11 antibodies.
neurology.org/doi/full/10.12…
Tyler Hulett retweeted
1/ We are excited to share our new report in @JExpMed showing that almost 40% cases of herpesvirus (HSV)-triggered fulminant viral hepatitis (FVH) are due to pre-existing autoantibodies neutralizing type I IFN (autoAbs-IFN) (doi.org/10.1084/jem.20251760).
Tyler Hulett retweeted
@DrBetofMDPhD #Melanoma immunotherapy expert @Stanford gives brilliant defense of TIL vs TCR-T or CAR-T cell therapy/Concept is #TIL are polyclonal and can better deal with variants: see slides/+ patients with CR have 96% 10 yr disease-free survival (NCI Surgery Branch data).
