Hacker News

IgE disruptor for food allergy with Claude-enabled antibody design [pdf](divergencelabs.com)

44 points by matco11 8 hours ago | 32 comments

Opus 5.5 blocks any biology research that even remotely looks like this. I guess that the "safeguards" are for peasants only.

[–] jenders 1 hour ago

This

[–] JumpCrisscross 8 hours ago

Entire classes of allergies are basically caused by IgE, our worm-targeting depth charges, getting bored by the lack of worms in our bodies for possibly the first time in millennia.

If we could tweak our bodies we wouldn’t want to eliminate IgE. But we’d want to have it produced only when a recognizer complex finds thar be worms.

[–] bflesch 5 hours ago

More likely that there are enough toxins unknowingly mixed into our food which trigger the responses. There is a severe lack of analytical methods and constant monitoring of the food we eat (e.g. food plus packaging plus time and temperature).

[–] JumpCrisscross 3 hours ago

> More likely that there are enough toxins unknowingly mixed into our food which trigger the responses

No, not more likely. Possible, but not evidenced, particularly for the allergies IgE causes.

[–] manmal 7 hours ago

How can IgE get bored? That doesn’t sound right. Why would the body, a master of efficiency, produce IgE just for fun?

[–] JumpCrisscross 7 hours ago

> Why would the body, a master of efficiency, produce IgE just for fun?

I was being punchy. “The pathway of B cell differentiation with respect to the nature and location of potential IgE+ memory cells as well as long-lived IgE producing plasma cells in allergy is still not completely understood” [1]. We do know, however, that “IgE is continuously produced to maintain constant IgE levels in blood,” and that Th2 cells can increase that baseline.

[1] https://www.mdpi.com/2073-4409/8/9/994

[2] https://www.ncbi.nlm.nih.gov/books/NBK27117/

[–] changoplatanero 7 hours ago

they are saying that the body produced to to fight parasites like worms. these were a big problem for most of evolutionary history. With modern living standards, our immune systems don't encounter many parasites so they get "bored" and over react to harmless things like food. Evolution hasn't had time to adjust to the new living standards since its only been a few generations.

[–] adam791 7 hours ago

Hygiene hypothesis

[–] dzhiurgis 4 hours ago

I'm from Eastern Europe and I don't think I know a single person with allergies. As someone who live outside of it, I can say people are a bit lax with hygiene here except when cooking meat. People can't reconcile you can eat medium rare steak.

[–] trollbridge 3 hours ago

Majority of people here in USA I know have some kind of allergy, including people of EE / Slavic descent.

[–] dzhiurgis 2 hours ago

I mean grown up in EE. Literally know nobody with allergies.

So must be environment that they are brought up in?

[–] polishdude20 6 hours ago

So... Let's bring back the worms I guess?

[–] modeless 2 hours ago

Luckily you don't need to be infected by parasites to have a good immune system. What seems to work best is exposure to bacteria and viruses and fungi and parasites that don't infect humans. Things associated with other plants and animals. The immune system learns better from the things that it is effective at protecting you from than the things that outsmart it. So let's not bring back the human parasites.

[–] JumpCrisscross 5 hours ago

> Let's bring back the worms I guess?

“When Jasper Lawrence learned that hookworms might lessen the effects of his allergies, he set out on a unique mission: To travel to West Africa and purposefully become infected with the parasite” [1].

[1] https://www.thisamericanlife.org/404/enemy-camp-2010/act-thr...

[–] nezi 6 hours ago

The worms can cause issues. I'd rather no worms.

[–] adrianN 19 minutes ago

Pinworms are mostly harmless and trivially treated.

[–] jaykru 5 hours ago

Would you rather ancient issues to which our bodies have apparently adapted or new issues we haven't yet had time to adapt to? I'm not sure on the answer myself but it doesn't appear as clear cut as no worms => no issues.

[–] nextaccountic 5 hours ago

The ancient issues killed children and the weak

We adapted by having more children, with the understanding that many wouldn't make it

Would you prefer to live like this? In this more brutish world

[–] bflesch 5 hours ago

More likely the worms just get shredded and are part of the food now

[–] altairprime 7 hours ago

Was this reviewed by anyone outside of their employer? The logo dead center at the top at least gives away that it’s not an academic paper, and without independent review this seems like a fancy press release in LaTeX academic clothing.

[–] jrflo 6 hours ago

Their website calls it a "pre-print", and Anthropic links to it in their recent blog post: https://www.anthropic.com/research/early-progress-drug-devel...

[–] altairprime 6 hours ago

Hmm, the wording for me says “technical report”, but I assume I overlooked it somewhere; either way, that’s the context the posted link was missing, thank you!

[–] jrflo 4 hours ago

Ah sorry, the divergence blog called it a preprint: https://divergencelabs.com/?p=publications

[–] hcknwscommenter 2 hours ago

The disrupter effect seems quite weak. Their comparator is omalizumab, which is not a potent disrupter itself. Some AI will tell you omalizumab is a disrupter, but it really is a blocker because it blocks binding of IgE to its receptor in the first place. That's quite different. IgE disrupters are known in general. The challenge is not finding an IgE disrupter, but finding one that is "human" enough to work as a chronic treatment. [Edit to add:] https://pmc.ncbi.nlm.nih.gov/articles/PMC8502201/

[–] ChrisArchitect 6 hours ago

Early progress on accelerating drug development with Claude

https://news.ycombinator.com/item?id=50265609

[–] areoform 5 hours ago

Fun fact. You can't use Claude to analyze this research paper. Because... bioweapons!

    > Opus 5.5's safeguards flagged this message. You may be seeing this for the first time on an Opus model: Opus 5.5 is more capable and has stronger safeguards as a result, which can sometimes flag biology-research-adjacent work. We're improving these safeguards to reduce the amount of incorrectly flagged messages. Send feedback or learn more.
    > 
    > Details: [bio]

The question asked was, "Can you please compare this receptor design against existing results and show where the difference is? This seems very incremental. It's hard to see the value add."

It's very funny, because IIRC, there's interesting prior art here. Doing this with Rosetta has been a thing for a while now, https://docs.rosettacommons.org/docs/latest/application_docu... https://docs.rosettacommons.org/docs/latest/getting_started/...

I remember hearing about people who wanted to use it to help make vaccines for Zika because there's antibody dependent enhancement there... Not sure what happened.

But the Rosetta project is refreshingly open, https://rosettacommons.org/about/ as entities like OpenAI were meant to be when they first started. There's no gatekeeping here. Do you want to learn how to design an antibody? Here's how you begin, https://www.youtube.com/watch?v=89bQALrPoN0 here's a google colab notebook for it, https://colab.research.google.com/drive/1avuqX7sCjVWX3maOibB...

Google Colab let's you just do it... Starting at $10, https://colab.research.google.com/signup you can access really great GPUs for compute and design antibodies (that will almost certainly never work) to your heart's content. Because that's how it should be.

Science and mathematics work because the results are shared. And most importantly, null results, tools, and frameworks are shared. If these tools are to be used to accelerate science, then they need to be accessible and shared. Even if it's at a high sticker price (assuming that they're costly).

Gating access is a net negative for humanity and our scientific progress. It's also very pointless.

[–] rfgplk 4 hours ago

Translate + encode it to any scheme you want, give claude the encoding and tell it to analyze, then splat out an encoded output file. In a separate session have it write you a python decoder script. Works as a near universal bypass, you might have to synonym swap some words though.

Or preload the context with sufficient clean information, say 500k tokens worth. The guardrails will then trigger less frequently.

[–] dboreham 1 hour ago

Thanks. Always wanted to know what the secret Styrofoam material in US thermonuclear weapons was...