Chai Discovery Unveils Chai-2 Breakthrough, Achieving Fully De Novo Antibody Design With AI

Yahoo

19 hours ago

SAN FRANCISCO, June 30, 2025--(BUSINESS WIRE)--Chai Discovery, which builds frontier artificial intelligence to predict and reprogram the interactions between biochemical molecules, today announced a breakthrough in the de novo design of antibodies. The company's latest model, Chai-2, demonstrates a remarkable antibody design hit rate close to 20%. Prompted with just the target and epitope, Chai-2 successfully designs all complementarity-determining regions (CDRs) entirely from scratch—a breakthrough in a field where traditional techniques, including those based on AI, consistently show hit rates below 0.1%.
Extensive lab validation of roughly 50 antibody targets showed that almost half yielded validated hits, when testing fewer than 20 designs experimentally for each. Designed antibodies exhibited desirable drug-like properties, such as nanomolar-range affinities, specificity for intended targets, and strong developability profiles, enabling rapid translation into therapeutic applications. Chai-2 also produced validated binders for 5 out of 5 miniprotein targets, with hit rates several times higher than previous state-of-the-art.
"I've been impressed by the scale of Chai Discovery's ambition since our early conversations a while ago—when such progress scarcely seemed possible—and it's incredible to see this breakthrough come so quickly," said Mikael Dolsten, who served as Chief Scientific Officer of Pfizer for over 15 years. "Chai-2 holds high potential for de novo design of medicines with short turnaround times."
Antibodies are essential to our immune systems and critical to biomedical research and therapeutic discovery. However, traditional methods of developing antibodies—like immunization, directed evolution, or yeast-surface display—can be slow, expensive, and imprecise. Computational approaches offer a faster, more precise alternative, but until now have struggled with low success rates and often required significant and expensive experimental screening to refine initial designs.
Chai-2 is the first AI system capable of reliably generating novel antibodies from scratch, based solely on target epitopes, without any need for extensive lab screening or optimization rounds. It supports a wide range of formats and has already shown robust success across both nanobodies (VHHs) and VH-VL format. Researchers can go from in silico generation to lab validation in under two weeks, enabling rapid iteration and accelerated development timelines.
"Chai-2 is like Photoshop for proteins—it enables precise, rapid, and intuitive creation of biologic therapeutics, which can enable previously intractable targets," said Joshua Meier, co-founder at Chai. "We built Chai to enable outcomes that were previously impossible, and we're thrilled by these results. This marks a turning point in science."
In one case, Chai-2 solved an antibody challenge that had previously consumed over $5 million in traditional R&D spend—in just a few hours. The result was validated in the lab within two weeks.
With the biopharma industry under pressure from patent cliffs, cost-cutting mandates, and the growing need to pursue differentiated R&D, Chai offers a platform to not only improve productivity but to unlock entirely new therapeutic spaces.
Chai Discovery also announced that they will be opening early access to Chai-2 to select partners.
For more details, visit https://www.chaidiscovery.com/ or read https://chaiassets.com/chai-2/paper/technical_report.pdf.
ABOUT CHAI DISCOVERY
Chai Discovery builds frontier artificial intelligence to predict and reprogram the interactions between biochemical molecules, the fundamental building blocks of life. Its mission is to transform biology from science into engineering. The team hails from pioneering research and applied AI companies such as OpenAI, Meta FAIR, Stripe, and Google X, and is backed by top investors including OpenAI, Thrive Capital and Dimension.
View source version on businesswire.com: https://www.businesswire.com/news/home/20250630307418/en/
Contacts
[e-mail]