Reinforcement Learning Produces Innovative Protein Designs

Researchers have recently adapted reinforcement learning, a type of machine learning, to create a powerful protein design program. This method has been used to solve long-standing puzzles in protein science and has been found to be more effective at generating useful antibodies in mice compared to previous methods. Scientists have used this approach to create hundreds of AI-designed proteins in the laboratory, which have been confirmed to be of an atomically accurate design. This is an exciting step forward in the field of protein design and could open up a new avenue for drug discovery.