Keep pulling the thread on Jack Dent and Joshua Meier.
The Chai 2 model maintained a consistent success rate even when tested on targets with as low as 25% sequence similarity to its training data, indicating strong generalization capabilities.
Jack Dent predicts that Chai Discovery's antibody design success rate could increase from 20% to over 50% or even near 100% within the next year.
Chai Discovery believes it is now feasible to generate entire drug candidates in a zero-shot manner directly from its models.
Using the Chai 2 platform, Chai Discovery demonstrated the ability to design functional antibodies with a success rate of approximately 20% from just 20 attempts per target.
Chai Discovery's Chai 2 antibody discovery platform can design antibodies that bind to specified targets with a 100-fold higher hit rate than previous computational approaches.
Jack Dent predicts that within the next few years, humanity will gain the ability to engineer molecules with atomic precision.
Chai Discovery's models achieved a 70% success rate in designing mini-proteins with picomolar affinities against all five targets tested.
In a partner collaboration, Chai Discovery's model designed a cross-reactive antibody that bound to both human and monkey (cyno) protein targets from a batch of only 14 sequences.
Chai Discovery's initial company-wide goal for its antibody design project was a success rate of only 1%.
The Chai 2 model can predict protein structures with an error resolution of less than the width of a single atom.
Chai Discovery will be opening up access to its Chai 2 platform for both academic groups and industry partners.
Previous computational antibody design techniques had a success rate of approximately 0.1% or lower.