Machine learning-assisted directed evolution of plant Rubisco.

Machine learning-guided evolution improves Nicotiana tabacum Rubisco's carboxylation efficiency by up to 43%, surpassing natural sequence diversity.

• Why it matters: Enhancing Rubisco's efficiency is crucial for increasing crop productivity and addressing global food security, yet engineering this enzyme has been historically difficult.
• What they did: Using the ML model ESM-IF1, researchers identified key amino acid sites in Rubisco, designed libraries of variants, and selected improved enzymes through growth in Rubisco-dependent Escherichia coli.
• The result: The study produced variants with novel amino acid changes, notably T391I, which increased carboxylation rate by 29% and efficiency by 43%, demonstrating ML's potential to explore new functional sequences.