Nature paper reports an artificial intelligence model that designs proteins from plain language descriptions

27 October 2026

A study in the journal Nature introduces Pinal, a foundation model with 16 billion parameters that designs candidate proteins from natural-language descriptions of a desired function. It was trained on 1.7 billion synthetically annotated protein-text pairs, linking functional intent to protein sequence and structure. The authors applied it to four design tasks - a fluorescent protein, a polyethylene terephthalate hydrolase, an alcohol dehydrogenase and a metabolic H-protein - and observed the intended function in each case, including catalytic activity for both designed enzymes. Without iterative experimental optimisation, a designed H-protein performed 1.7-fold better than the corresponding native H-protein in a multi-enzyme carbon dioxide fixation pathway.