Anthropic has chosen Twist Bioscience as one of its independent evaluators in testing whether Claude can help accelerate parts of the drug design process by using the artificial-intelligence model to design minibinders for multiple protein targets.
Minibinders are small proteins designed to latch onto a target protein. Binding is how many modern medicines work; the minibinder attaches to a target and inhibits, activates or delivers something to it.
Designing a new binder has historically taken protein engineers months of computation, optimization and screening per target, but machine-learning models could rapidly expedite the process, Anthropic said Wednesday.
The AI company conducted a multi-arm protein design campaign against 15 targets using its Claude Opus 4.8 and Mythos Preview models. Twist Bioscience and Adaptyv Bio are the independent evaluators checking Claude's work.
Twist Bioscience and Adaptyv Bio independently produced and tested Claude's designs in the lab and found that, of the 15 targets Anthropic designed against, Claude successfully designed binders against 14 of them.
"These include high-affinity binders against at least six targets, and binders matching or exceeding the best reported affinity against at least four targets," Anthropic said. (Affinity is a measure of how strongly a protein binds to its target, with high-affinity binders generally needed to achieve a therapeutic effect.)
Anthropic said the experiment shows promise that Claude can help protein designers accelerate parts of the drug design process.
Comments