Claude 参与设计蝎子抗毒蛋白，17 个设计通过结合测试

🦂 Claude just helped design a potential scorpion antivenom. In 64 hours, an internal research model generated over 318,000 protein designs, guided by five human messages. Anthropic’s scientists and Citrini Research sent 211 designs to outside labs for testing and 17 met the study’s binding criteria. These tiny proteins grab scorpion toxins, potentially blocking their attack on nerve cells. But binding isn’t the same as neutralizing venom… The team hasn’t tested whether they protect cells or animals, and each successful design bound only one toxin. It’s an early research result with a much bigger question ahead. Can these molecules eventually become safe and effective medicines? Research paper: “De novo design of protein binders against scorpion toxins with Claude”

AI 导读

Anthropic 科学家与 Citrini Research 用内部研究版 Claude 在 64 小时内生成超过 318000 个蛋白质设计，全程仅由 5 条人类消息引导。

打开原文tiktok.com

选题角度

原文给出了64小时、318000个设计、211个送检、17个达标这组数字，适合做热点评论，讲清结合毒素与中和毒液之间的差距。

博主

Rowan Cheung@rowancheung

粉丝 8.4万

The latest in AI and robotics, explained Founder & CEO of The Rundown AI Zurich 🇨🇭 / NYC 🇺🇸

互动数据

浏览: 1313
点赞: 48
评论: 2
分享: 2
收藏: 2

数字来自平台，为最近一次抓取时的数据。

正文 · 原文

以下是视频口播的自动转写，可能有错字。

So in case you missed it, clog just helped design a potential scorpion antivenom. In 64 hours, it generated 318,000 protein designs guided by just 5 human messages.

Outside labs then tested 211 of those designs, and 17 passed the study's binding tests. For context, these are tiny proteins designed to grab the part of a toxin that attacks your nerves.

See, scorpion neurotoxins interfere with sodium channels, the molecular gates that help nerve cells fire. And that disruption can trigger a dangerous surge of stress chemicals, and in severe cases, heart failure.

The idea is to cover the toxin's docking site before it can reach those channels. The project actually brought together investment researchers from citrine and anthropic scientists using an internal version of clog and specialized design tools.

But we still don't know whether grabbing these toxins actually stops them. See, it's still really early, and nobody has actually tested it in cells or animals yet.

Each successful protein only bound one toxin, so they're still far away from the universal antivenom. But still, these designs made it out of a computer and into a lab where scientists confirmed they could grab their target.

But the next challenge is actually finding out whether any of them can help someone who's been stung in the real world. If you want to stay in the loop with the latest in AI and robotics, follow for more.