Predicting how proteins work now takes 600 times less data and computing power

A new AI model can predict how changes to proteins affect their function. It does this using much less data and computing power than previous methods, making protein engineering significantly cheaper and faster to develop.

For years, designing new proteins or understanding mutations meant expensive lab experiments or huge computing clusters. This paper shows a way to get similar results with a fraction of the resources. This means smaller labs or companies can now afford to experiment with AI-driven protein design, potentially speeding up the search for new medicines or industrial enzymes.