Cradle

We're hiring for our Large Libraries team in Amsterdam! 🧬 Head of Large Libraries 🧬 (Senior) Research Associate (new) 🧬 (Senior) Scientist (new) Cradlers are all deeply motivated to further our mission: helping researchers engineer better proteins faster. Our Large Libraries team is key to this, developing methods to generate massive datasets (>10^6 data points) to power our platform, ultimately helping our users get to their target protein quicker. We've been building out these methods for a while, and now we're ready to take things to the next level. We're currently expanding our lab space and looking for enthusiastic researchers to join us. If you're excited about FACS, microfluidics, and/or display platforms (yeast, phage, mRNA), please apply!

I've watched companies spend more on their Phase 1 trial catering budget than they were willing to invest in improving the molecule going into that trial. A slightly better lead candidate can literally mean a $2B difference. And the crazy thing is that R&D teams understand this, but CFOs often don't. They see an investment in a new technology, along with experimental costs, and it's really an expensive line item. They make risk assessments, but rarely map opportunities. Ironically, the same companies will spend $50M on a Phase 2 trial for a candidate that's marginally better than what failed in Phase 1. Yet, spending money to engineer a better candidate upfront is the harder sell! I'm not saying finance is wrong to scrutinise costs. But when you're in a business where a single approved drug generates billions, optimising mainly for risk profile instead of candidate quality is solving the wrong problem, isn't it? Am I wrong here?

We were tired of waiting for our DPO trainings to finish overnight… so we built g-DPO. Now my preference optimisation runs finish over my (admittedly French 🇫🇷😉) lunch breaks. I’m excited to share that our work on scaling preference optimization for sequence models has been accepted at two (N)eurIPS workshops ! Come check out our posters: 🇩🇰 MLSB (Copenhagen) EurIPS — Dec 7 🇺🇸 FM4LS (San Diego) NeurIPS — Dec 7 I’ll be in 🇺🇸 for NeurIPS main conference and in 🇩🇰 for the workshops. Looking forward to catching up with familiar faces and meeting new ones ! Preprint here: https://lnkd.in/dWRNMB38 Jonathan D. Ziegler Nicolas Deutschmann Arthur Lindoulsi Eli Bixby Cradle

At Cradle, we need to use every tool in the protein engineering toolbox for high-throughput protein engineering with high data quality and quick turnaround times. Protein tags are one such tool. These molecular "handles"—some just 6 amino acids in size—confer useful properties to our proteins that accelerate our lab workflows and help us validate our machine learning modeled sequences faster and more reliably. In this article, we're breaking down what protein tags are, how we use protein tags in our lab, and which ones are our favorites.

The track record for software in pharma isn't great. But a few things have fundamentally changed. I got this comment under a post a few months ago regarding Biotech vs Tech VC, and it's legit pushback. First of, there are successes. Veeva Systems , Benchling , Dotmatics , Tempus AI. Veeva alone has a PE ratio of 59 with $45B in software revenue. Current public companies are lagging indicators - there's more coming. I also see these 4 shifts: 1) Pharma's willingness to buy vs. build has increased. Across cloud/infra services, data analysis/workflow tools, and application software, spend is rising. When I was at Google 4 years ago you saw hardly any large pharma contracts. SaaS allows agile experimentation, quick deployment, and lower upfront risk. 2) Companies are tapping research and clinical budgets. Axiom, Automata, Kaleidoscope - they're pulling from budgets typically reserved for non-software. Market pressure. Pharma feels the risk of margin compression (China, FDA, NIH, patent cliff) and is trying to move faster. 3) Tech/AI is becoming more useful. The magnitude of impact some tools can have has increased in the last couple of years. 4) Tech VCs are getting more dangerous. Firms like GV and Dimension have intimate knowledge of both tech and bio. Software in pharma is still hard. But the conditions for building real businesses have changed. Thoughts?

Pharma CFO: "What's the ROI on this protein engineering AI thing?" Me: "Depends. What's your lead candidate worth if it's 15% better?" CFO: "Define better." Me: "Higher binding affinity, better expression yields, improved thermostability." CFO: "In dollars." Me: "If higher binding reduces your Phase II failure risk even a few points, that’s tens of millions in avoided sunk cost. If higher expression means 2× output per bioreactor run, that’s a 5–15% cut in COGS, or $100–300M over the product lifecycle. If better stability reduces cold-chain or opens new markets, that’s potentially hundreds of millions to billions." CFO: "That's a lot of 'ifs'." Me: "It is. So is every drug development program." CFO: "I can't put 'potential billions' on the balance sheet." Me: "Right. Which is why you’ll end up with a molecule that’s only incrementally better... and still spend $500M to push it through trials." CFO: "As opposed to?" Me: "Investing now to get a molecule that’s materially better? and spending the same $500M on a candidate with a real shot at being best-in-class." CFO: "You're saying this improves our odds." Me: "I'm saying this is the cheapest part of your development budget and has the highest leverage on outcomes. But it shows up in this year's P&L, so it feels expensive." CFO: "...I'll think about it." Want to learn how scientists engineer better proteins faster with Cradle? Visit: https://lnkd.in/et3wNHYY

#EurIPS: Join Cradle and Adaptyv for drinks and conversation at the intersection of machine learning and protein engineering. Whether you're working on foundation models for biology, optimizing antibodies, or exploring the latest in generative protein design. Come connect with others pushing the boundaries of ML×Bio. What you can expect: - Casual networking with researchers and practitioners in computational biology - Discussions on the latest in protein language models, structure prediction, and AI-driven protein engineering - Drinks and good company No formal program. Just an opportunity to meet the community building the future of biological engineering. Co-hosted by Cradle and Adaptyv Bio during EurIPS week in Copenhagen on December 6. Space is limited. RSVP to secure your spot: https://luma.com/e0avnsib

We're #hiring a new Head of Large Library Generation in Amsterdam, North Holland. Apply today or share this post with your network.