Synthetic Manihot esculenta Rubisco activase proteins with increased thermotolerance identified via machine learning

Adaptation to increasing environmental temperatures is essential to plant survival and human food production. Thermal tolerance is controlled by a complex network of factors in plants including but not limited to genetic variation and environmental context. Rubisco activase (Rca) is a key photosynthetic enzyme with low thermal tolerance. Here, we report a large machine learning-directed screen of >1,400 synthetic cassava Rca enzymes which identified mutations that convey increased thermal stability while minimizing introduced mutations. We demonstrate multiple synthetic proteins that maintain activity at 8°C higher than wildtype cassava Rca including a single mutation that retains most activity post heat-shock.