OpenAI and Retro reveal GPT-4b micro, a model for engineering proteins used on reprogramming factors

Retro Biosciences/GPT-4b micro (with OpenAI)

OpenAI and Retro Biosciences revealed GPT-4b micro, a specialized language model trained on protein sequences for predicting protein engineering improvements. Researchers used its suggestions to redesign two Yamanaka factors, making them more than 50 times as effective at preliminary measures, demonstrating the model's first application to longevity science.

• GPT-4b micro is specialized language model for biological data; trained on protein sequences and interaction data, not general-purpose AI.
• Suggested modifications made two Yamanaka factors >50x more effective at cell reprogramming on preliminary measures.
• Model frequently proposed changes to approximately one-third of amino acids in these factors; outside verification pending peer review publication.