Machine learning streamlines the complexities of making better proteins

Scientists developed MULTI-evolve, a machine learning framework that predicts how proteins will perform when multiple amino acids are changed simultaneously, eliminating the need for countless rounds of trial-and-error testing. The system combines laboratory experiments with AI to identify optimal combinations of amino acid swaps in just one testing round, rather than the traditional iterative approach. Testing on three proteins, including antibodies and CRISPR components, successfully produced enhanced versions that outperformed originals, potentially revolutionizing protein design for medicines and biotechnology applications.