TLDR: Chai Discovery has launched Chai-2, an advanced AI model achieving a remarkable 16-20% experimental hit rate in de novo antibody design, a 100-fold improvement over previous computational methods. This breakthrough allows for the creation of novel antibodies from scratch in as little as two weeks, significantly accelerating drug discovery and opening doors to previously ‘undruggable’ targets.
SAN FRANCISCO – Chai Discovery, a leader in applying artificial intelligence to molecular design, has announced the release of Chai-2, a groundbreaking AI model poised to transform the landscape of de novo antibody and binder discovery. The new model boasts an unprecedented experimental hit rate of 16-20% in de novo antibody design, a significant leap forward that represents an over 100-fold improvement compared to prior computational methods which typically yielded hit rates below 0.1%.
Traditional antibody discovery methods, such as animal immunization or large-scale library screenings, are notoriously slow, expensive, and often struggle against challenging or novel targets. Even previous computational approaches, while promising efficiency, still necessitated extensive experimental screening due to their low success rates. Chai-2 addresses these bottlenecks by offering a fully zero-shot generative platform that bypasses the need for traditional high-throughput screening.
Chai-2 combines all-atom structural modeling with a generative design model to produce antibody sequences entirely from scratch. Unlike conventional methods, it does not rely on existing antibody templates or large screening libraries. Instead, it designs complementarity-determining regions (CDRs) based solely on a given epitope and target structure. This innovative approach enables the generation of new sequences that differ significantly from known antibodies, making it applicable to difficult target sites.
The model’s efficacy has been rigorously validated. In tests involving 52 previously unaddressed antigens, Chai-2 achieved its double-digit hit rates by evaluating only up to 20 candidates per target in a standard 24-well-plate format. Remarkably, in a single experimental round, at least one successful binder was discovered for 50% of the tested targets. The designed antibodies exhibited highly desirable drug-like properties, including nanomolar-range affinities, high specificity for intended targets, and robust developability profiles, facilitating their rapid translation into therapeutic applications.
Beyond antibodies, Chai-2 also supports the design of various other modalities, including single-chain antibodies (scFv), VHH nanobodies, and miniprotein scaffolds. For miniproteins, the model demonstrated an impressive 68% wet-lab hit rate across five test targets, routinely yielding picomolar affinities.
One of the most transformative aspects of Chai-2 is its speed. The entire design-to-validation cycle, from AI generation to wet-lab characterization, can be completed in under two weeks. This dramatically compresses timelines that traditionally spanned months or even years. In a notable instance, Chai-2 solved an antibody design challenge that had previously consumed over $5 million in traditional R&D spend, achieving a validated result in just a few hours.
This breakthrough marks a pivotal shift in drug development, moving from empirical exploration to deterministic molecular engineering. Industry experts believe Chai-2’s high success rate and broad applicability offer the potential to tackle traditionally ‘undruggable’ targets and accelerate the development of complex therapeutic forms like antibody-drug conjugates and bispecific antibodies.
Chai Discovery, founded in early 2024, quickly established itself with its first model, Chai-1, which achieved a 77% success rate on the PoseBusters benchmark without relying on computationally expensive multiple sequence alignments. The company has attracted significant investment, raising nearly $30 million at a $150 million valuation from prominent investors including OpenAI, Thrive Capital, and Dimension. The Chai team comprises veterans from leading AI and tech companies such as OpenAI, Meta FAIR, Google X, and Stripe.
Chai Discovery plans to open early access to Chai-2 for select academic and industry partners, prioritizing projects that aim to benefit human health and society.
