A team of researchers has developed an AI model capable of generating the code to synthesize novel proteins, including a previously unknown bright fluorescent protein. The model, called ESM3, was trained on massive amounts of data, including 771 billion tokens from protein sequences and structures, allowing it to mimic the evolutionary process of a protein that never existed naturally. After training for virtual time, the AI generated a new green fluorescent protein, esmGFP, which has a unique genetic sequence different from other known proteins. The researchers suggest their model could be used to create new proteins for various applications, including medicine and environmental research.
Source: https://phys.org/news/2025-01-ai-simulates-million-years-evolution.html