Introducing Claude for Life Sciences: Streamlining Drug Discovery

View profile for Dave Nolan

Health AI @ Anthropic

Launching: Claude for Life Sciences 🧬 🧪 🔬 🤖 🥼 The bottleneck in drug discovery is execution overhead. Researchers spend most of their time formatting, searching, and standardizing instead of thinking and testing. Here's what we're changing: Claude now connects directly to the tools scientists already use. Benchling for protocols and experimental records. PubMed for literature. 10x Genomics for genomic analysis. BioRender for figures. Instead of jumping between platforms and manually synthesizing information, scientists ask Claude to pull what they need and generate what comes next. We also built Agent Skills—basically lab protocols that Claude follows consistently. First one handles quality control for single-cell RNA sequencing using scverse best practices. More coming - stay tuned!! The system effect: Less time on execution overhead means faster iteration cycles. Schrödinger reports 10x speed gains on specific tasks. Sanofi has most employees using Claude daily. 10x Genomics enables researchers to run complex analyses without computational expertise. Claude Sonnet 4.5 now scores 0.83 on Protocol QA versus human baseline of 0.79. It understands laboratory procedures at expert level. Faster cycles compound. Breakthroughs arrive months sooner. Full details: https://lnkd.in/g47HTaAs If your team is spending more time formatting than discovering, let's talk about your specific workflow challenges. Very excited to speak with you!

Daniel H.

Founder, ML engineer

1w

I've been working with Claude Code Pro for 6 months now and it can barely handle a 20-25k repo (that's just simple backend/frontend w maybe 3k of ML code). It forgets things all the time, it disregards instructions, it writes boilerplate no one asked for, it makes elementary mistakes disregarding in-context information that it pretended to know 2 seconds ago. The reason I can still use and make use of Claude is because this is a 25k line repo that I know like the back of my hand and I can catch Claude the moment it does something stupid (which it does at least 3-5 times a day). None of this applies to a researcher's work. I happen to know a bit about bioinformatics and biomedical research and I can guarantee that this product (and anything built on today's LLMs) will lead to some horrendous "science" and a ton of made up junk polluting our scientific knowledge bases even further. I get it why ppl will jump on it (research is hard, often boring and underpaid) but the current tech is just not good enough for this. Like not even remotely and I'm not alone w this opinion https://www.youtube.com/watch?v=lXUZvyajciY

Jason S.

AI x Bio | Executive | Author

1w

Several months ago I was curious how hard these agents were to build. Turns out they were pretty straight forward. I describe a few here that perform well on things like scRNA Seq and LabBench tasks that were coded up as an experiment. https://biotechbio.substack.com/p/biomedical-ai-agents

Michael Raspuzzi

Worldwide Studios | We help people level up their skills in AI, robotics, and applied sciences

1w

Luis 👀 we may need to do a claude skills lab auto hack... Eslam Elshahat, Ph.D. great to see Benchling as one of the launch partners, congrats!!

I hope Anthropic rapidly expands capabilities from Pre Clinical also into Clinical Development, with Protocol to Trial Design Optimization, Eligibility Optimization, Modernization of Clinical Data Flow aligned with CDISC 360 . What problem solvers in clinical development needs.

Eric Luellen

CEO | CTIO | CDSO | AI/ML in Translational Medicine | TechBio | FDA-Designated Therapies | 3x Founder | 1x Exit | Strategy & Innovation | American Univ. | Northwestern | Harvard

1w

But you could already do most of this by simply uploading the relevant artifacts as PDFs.

Daniel Uribe, MBA 🧬🔗🔐

CEO & Founder, GenoBank.io 🧬| Making Web3 Genomics | Inventor of BioNFTs™ (2 USPTO Patents) | PhD Candidate, Decentralized Biobanking | Techstars Web3 ’24

1w

Congratulations Dave! This is exactly why we built BioFS - a BioNFT-gated filesystem where researchers actually own their genomic data while accessing Claude-powered analysis. (MCP server) The execution overhead you're solving perfectly complements what we're doing with BioIP™ on Story Protocol. Imagine: researchers query their tokenized genomic datasets with a prompt through Claude, then instantly license that data as programmable BioIP - like copyright for DNA bioinformatics derivatives We're seeing the same 10x gains when scientists stop wrestling with command lines and start asking "which variants correlate with treatment response?" Their data becomes a sovereign asset they control, not another corporate dataset. GenoBank.io BioFS 🧬 + Anthropic Claude + Story = researchers own, analyze, and monetize their discoveries in one flow. The bottleneck isn't just execution - it's that scientists don't capture value from their own breakthroughs. Tokenization changes everything. 🧬⚡ Would love to explore how this plugs into your ecosystem! https://genobank.io/blog/beyond-data-breaches-biofs-claude-story-protocol-licensable-bioip.html

Shloki Jha

Applied AI & ML | NLP & LLMs | Predictive Analytics | Ex-Rocket Companies | MSDS UofM ’25 | CE IITR ’23

1w

Exciting work! I’ve been exploring Claude’s reasoning on multimodal data, and it’s remarkably perceptive compared to other models, looking forward to seeing its impact on life sciences workflows.

Michael Unger

GTM Strategy & Growth | 0→1 Product Commercialization | DeepSky (acq. by Airtable) | AI Go-to-Market Leader | Oxford Alum | ex Founder, Investor

6d

Fah Sathirapongsasuti check this out 👏🏻

See more comments

To view or add a comment, sign in

Explore content categories