/ai

Anthropic's Claude Designs Protein Binders for 14 Targets · Digg

Anthropic's Claude Designs Protein Binders for 14 Targets

Anthropic reports Claude designed viable protein binders for 14 of 15 targets using expert prompts.

TLDR

Anthropic announced that Claude used expert-written prompts to run protein binder design workflows and produced functional binders against 14 out of 15 biological targets. The company stated the effort replaced weeks or months of traditional expert work per target and released technical reports plus a video of nine experimentally confirmed binders. Replies note that specialized open-source models performed the core design steps rather than Claude alone and that the reported success rates match those of existing tools such as BindCraft for similar targets.

Featured Source

Combined views

4.6M

34 Sources, first seen 37d ago

21.2K likes

TagsAnthropic Claude Opus 5

Today's Rank

—

Not ranked yet

37d ago

first seen 37d ago

Anthropic's Claude Designs Protein Binders for 14 Targets

Anthropic reports Claude designed viable protein binders for 14 of 15 targets using expert prompts.

TagsAnthropic Claude Opus 5

34 Sources, 37d ago, first seen 37d ago

TLDR

Anthropic announced that Claude used expert-written prompts to run protein binder design workflows and produced functional binders against 14 out of 15 biological targets. The company stated the effort replaced weeks or months of traditional expert work per target and released technical reports plus a video of nine experimentally confirmed binders. Replies note that specialized open-source models performed the core design steps rather than Claude alone and that the reported success rates match those of existing tools such as BindCraft for similar targets.

Featured Source

Today's Rank

—

Not ranked yet

Combined views

4.6M

34 Sources, first seen 37d ago

21.2K likes823 comments5.5K saves2.6K reposts

Sentiment

Positive——Negative

Summary

Not enough discussion yet.

No sentiment analysis available yet.

823 comments

5.5K saves

2.6K reposts

Sentiment

Positive——Negative

Summary

Not enough discussion yet.

No sentiment analysis available yet.

219012345678901234567890123456789

219012345678901234567890123456789

219012345678901234567890123456789

219012345678901234567890123456789

34 Sources

@AnthropicAIMany drugs work by binding to a specific target in the body and blocking or changing what it does. An important first step in the drug development process is designing a molecule that can bind tightly to its target. Traditionally, that's meant weeks or months of expert work per target, sifting through a large number of candidates to identify the few that work. We wanted to test if Claude could successfully design novel protein binders from scratch (also called de novo design). With a protein design prompt written by a human expert, Claude autonomously designed protein binders against 14 out of 15 targets. We then worked with Adaptyv Bio and Twist Bioscience, who independently built and tested the proteins Claude designed.

@repligatewow.... claude invents new forms of pasta shaped like each letter of the word "ANTHROPIC"

@nc_freyToday we’re sharing an update on Claude’s protein design capabilities. Working with Adaptyv Bio and Twist Bioscience, Claude designed de novo protein binders against 14 of 15 targets. With a 30k token prompt written by a human expert, Claude achieved hit rates up to 28.2% when designing against all targets simultaneously over 48 hours, with no human intervention or steering. When designing against a single target in a 24 hour campaign, Claude achieved an overall hit rate of 35%. We included all targets in Adaptyv’s BenchBB, and novel targets 15-PGDH and GDF-8. Claude performs at or beyond the level of the top competition participants in both hit rate and binding affinity.

@rohanpaul_aiAnthropic published one of its most ambitious scientific experiments with Claude yet. Anthropic showed that Claude can take a biological target and autonomously run the computational protein-design campaign needed to produce binders that actually work in the lab. A lab may no longer need a dedicated protein-design expert to manually run every computational step. This can move the bottleneck from designing and triaging thousands of candidates toward experimentally testing a much smaller, AI-selected set - Given a detailed expert-written protocol, it researched each target, chose where to bind, installed and ran open-source protein-design tools, generated candidates, filtered and improved them, then picked the final proteins for lab testing. Humans did not make the individual design decisions. - The designs actually worked in the lab. Across 1,320 designs with usable measurements, 354 bound their intended targets, a 26.8% hit rate, and Claude found binders for 14 of 15 targets. - On several targets, its results were competitive with human/open design competitions. Claude had higher hit rates on 4 of 6 comparable competition targets. - Giving the agent more attention and compute seems to help. Mythos Preview reached a 35.1% hit rate when each target received its own 24-hour campaign, versus 26.7% when many targets shared a 48-hour campaign. - The AI still cannot reliably know when a whole campaign has failed. Some unsuccessful targets received computational scores similar to successful ones.

@AndrewCurran_I'm reading this now, but I wanted to make a post linking the associated reports and data so that anyone who wants to can discuss them with their model of choice. https://www.anthropic.com/research/Claude-accelerates-protein-design

@kimmonismusThis is interesting: Claude is already achieving roughly twice the protein-design hit rate of conventional human-led workflows. 27% hit rate in autonomous protein binder design, roughly twice the typical 10–15% rate reported in the field. Working from one expert-written protocol, Claude designed binders against 14 of 15 measurable targets. Independent labs confirmed that 354 of 1,320 designs bound successfully. Depending on the setup, Claude’s hit rate ranged from 22.6% to 35.1%. Its top-ranked design bound in 49% of campaigns. This is not yet fully autonomous drug research, but it is another important building block in that direction.

@NickADobosQuoted section is from this big boy https://huggingface.co/datasets/Anthropic/claude-protein-binder-design/blob/main/prompts/prompts/multi_target_binder_design_prompt.md Full repo of prompts used https://huggingface.co/datasets/Anthropic/claude-protein-binder-design/tree/main/prompts/prompts

@teortaxesTexAnthropic going all in on bio would be good Less time for Dario to contemplate cyberattacks… and hopefully he doesn't have the balls for ethnically targeted bioweapons instead. So we may genuinely get rapid progress in medicine.

@pdhsua nice demonstration of Claude Science, but worth clarifying that the design is not "done by Claude" but by orchestrating tool calls of open-source, task-specific protein design models: PXDesign, RFdiffusion, Genie, BoltzGen, etc I think the direction of LLMs using biology-specific models is a good one!

@EricTopol@AnthropicAI Yes, that's terrific. It will accelerate treatments, not cures. The latter are extremely hard to come by, but worth aspiring for.

Related

Researcher Questions Claude Opus Constitutional Midtraining

Theia Vogel questions whether hacker!opus received constitutional midtraining on certain language.

Anthropic Engineer Addresses Claude Opus 5 Complaints

Users report Claude Opus 5 feels inconsistent and lazy compared to version 4.6.

Debate Over Anthropic Claude Watermarking Technique

Posts analyze Anthropic's text watermarking technique and user reactions to it.