Chai Discovery Introduces Breakthrough in AI-Driven Antibody Design
Chai Discovery has announced a major advance in de novo antibody design with its latest AI model, Chai-2. The model achieves a design hit rate of nearly 20%, a significant leap from the sub-0.1% rates seen with traditional and AI-based methods.
Chai-2 generates all six complementarity-determining regions (CDRs) entirely from scratch, using only the target and epitope as input. It requires no template structures or multiple optimisation rounds.
In lab tests on around 50 antibody targets, nearly half produced validated hits after evaluating fewer than 20 designs each. The antibodies showed nanomolar affinities, high specificity, and strong developability.
The model also performed well on miniprotein targets, delivering hit rates well above previous best-in-class methods.
It supports a range of formats, including nanobodies (VHHs) and paired VH-VL chains. Designs can be validated in under two weeks from initial generation.
Traditional antibody development methods—like immunisation or yeast display—are slow, costly, and often unreliable.
Earlier AI tools sped up design but still needed extensive lab screening. Chai-2 overcomes these challenges with high success rates and minimal lab work.
At a time when the biopharma industry faces rising pressure to cut costs and innovate, Chai-2 offers a powerful new tool to speed development and unlock previously unreachable targets.
