JURA Bio, Inc.

So pleased our work on #VariationalSynthesis was selected as one of the top 4 papers at MIT Jameel Clinic - AI & Health's 5th Molecular Machine Learning Conference, #MoML2024. Eli Weinstein will be giving a short talk in person if you'd like to learn more live. Early bird registration is already sold out, but you can get regular tickets or onto the waitlist at https://moml.mit.edu. And you can always reach out to us if you have to miss the in-person event but would still like to learn more about manufacturing-aware generative protein models. JURA Bio, Inc.

It was a blast weighing in foundation models and their promise for biological sequence design as part of this panel and the Genetic Engineering & Biotechnology News State of AI in Drug Discovery summit. Hope you'll register (free) and tune in next week: https://lnkd.in/edNnQrW5

As an R&D stage company, at JURA Bio, Inc., we wanted to work out a way to estimate clinical efficacy before we entered the clinic. To do so, we turned to the only data sets available large enough to let us make good estimates about novel therapeutics and against unseen targets: the natural experiments that human adaptive immune systems undergo every day they see an infection or beat back an aberrant cancer cell. The result is CAIRE, released today, which can turn data about these natural experiments into a tool for amino acid-level therapeutic design before a discovery campaign is run. This paper presents a new approach to bringing predictions of efficacy into the earliest stages of the therapeutic design process. It has implications for cancer vaccine design, for TCR Ts, TCR mimicking antibodies and for soluble TCRs. Hope you'll check it out here: https://lnkd.in/e8H6aFQB and read more in our blog post: https://lnkd.in/esAzSryU

Variational synthesis is a new generative design paradigm for biology 🧬 out of JURA Bio, Inc. It bridges a huge gap between digital design of biological sequences (proteins, DNA, etc.) and physical manufacturing—at scale. By *learning the optimal ratios of available sequence templates* needed to recreate a target sequence distribution, variational synthesis learns the "recipe" to manufacture large libraries of biological sequences in a single pooled experiment In my writeup, I lay out the intuition, walk through the methods, results, and the talk about the bigger picture https://lnkd.in/gnVVRj2P

⏰ In our newly released preprint https://lnkd.in/e8KKUhmV (and accompanying blog post https://lnkd.in/eBiqw5UN), we introduced a novel class of generative AI with downstream translation in mind: 🔬 It effectively reduces the cost of gene synthesis by 1-trillion fold, and allows you to take your rigorously designed in silico proteins and put them directly into the gamut of functional testing. ➡ It's a win-win of AI-design meets high-throughput screening. 🙏 A big congratulations to the entire JURA Bio, Inc. team; we're eager to start delivering cures. Please don't hesitate to reach out if you're interested in learning more.

👨🔬 1 quadrillion dollars. 👨🔬 That's how much each of our AI-designed protein libraries would have cost to make using any other technology available today. #VariationalSynthesis is a profoundly new AI architecture that allows us to have our pick -- AI-designed proteins in the real world, ready for testing; AI-designed proteins on the computer, ready for in silico characterization; the same designs in the lab and on the computer with effectively no trade off in quality or scale. It's a set of generative models, built for designing new therapeutic antibodies, TCRs, and cell therapies; for generating target tissues at scale for disease insights; and for building important tools like polymerases and gene editors. And it's a physical, synthetic chip for sampling from generative models that effectively and reliably reduces the cost of gene synthesis by about 1-trillion-fold from trusted industrial gene synthesis providers. We're advancing insights into protein and therapeutic designs that previously could only be expressed on the computer and moving them into the real world, ready for testing. Artificial intelligence made real for biology. For those of you who have been long struggling with how to reconcile your understanding of drug development -- with all its unexpected twists and turns when observed in vitro and in vivo -- with the potential of generative AI and other machine learning methods for #BioAI and #TechBio for overcoming barriers in our understanding and design of new therapies, I am proud to announce our efforts at JURA Bio, Inc. to start putting those pieces together for you. I invite you to read more about it at our new blog post detailing the method: https://lnkd.in/eyF8wYSA as well as in our preprint: https://lnkd.in/eSxRhVeJ. At JURA, we're bringing generative AI to life, and we hope you'll join us. #TechBio #ArtificialIntelligence #BioAI