Unseen protein sequences outside the boundaries of natural biology can carry out functional molecular tasks. Scientists long assumed that billions of years of evolution had already mapped out most workable amino acid combinations. Instead, unmapped chains of amino acids physically link together and fold into active molecules that perform biological duties without matching known wild relatives.
A protein functions when its sequence of amino acid subunits bends into a precise three-dimensional structure. These molecular units lock together like teeth in a zipper, stabilizing the overall architecture of the molecule. The folded structure binds target chemicals or catalyzes biochemical reactions within a living cell. Computer scoring models, including protein language models, attempt to estimate this activity by evaluating sequence patterns, but their predictions struggle to track physical function.
Researchers assembled and evaluated the largest known collection of diverse natural orthologs alongside entirely new-to-nature sequences. The team measured the experimental fitness and activity of these diverse proteins across three distinct biological families. In each family, many synthetic sequences with low amino acid identity to existing natural proteins demonstrated strong biological activity.
These findings provide experimental evidence that natural evolution has sampled only a tiny fraction of all functional protein possibilities. The mapped functional space between existing proteins provides a foundation for engineers to advance artificial protein design and investigate evolutionary history.