These polypeptide letters and numbers could spell out a micron-scale poster or provide twee decor for a DNA origami bookshelf. They’re also a demonstration of the more practical capabilities of a generative artificial intelligence model for protein design. The researchers at biotech company Generate:Biomedicines who wrote the model provided it with virtual molds in the shape of letters and numbers. It returned amino acid sequences that could fold into the shapes shown, to fill these spaces. When researchers synthesized other proteins that the model had dreamed up based on prompts like the catalytic site of an enzyme, the proteins generally folded to match closely with the model’s prediction. Generate plans to use the model to design new enzymes and therapeutics for oncology and infectious disease treatment. Unusually for the field of protein-related AI tools, the company made the code underlying the model freely available (Nature 2023, DOI: 10.1038/s41586-023-06728-8).
Do science. Take pictures. Win money. Enter our photo contest here.