Optimized Machine Learning for Autonomous Enzymatic Reaction Intensification in a Self-Driving Lab

Optimizing enzymatic catalysis is crucial for enhancing the efficiency and scalability of many bioprocesses such as biotransformations, pharmaceutical synthesis, and food processing, as well as for improving the performance of analytical applications, including assays and biosensors. However, optimizing these reactions is challenging due to the multitude of interacting parameters such as pH, temperature, and cosubstrate concentration that require precise adjustment for maximum enzyme activity. Current optimization methods are often labor-intensive and time-consuming, especially when accounting for complex parameter interactions in highly dimensional parameter spaces. To overcome these challenges, we present a machine learning-driven laboratory platform that enables rapid, data-informed optimization of enzymatic reaction conditions in a fully automated environment. By conducting over 10,000 simulated optimization campaigns on a surrogate model generated via linear interpolation of experimentally obtained data, we identified and fine-tuned the most efficient machine learning algorithm for optimizing enzymatic reactions. This allows the platform to autonomously determine optimal reaction conditions with minimal experimental effort and without human intervention. The effectiveness of our approach is demonstrated by the accelerated optimization of reaction conditions in a five-dimensional design space across multiple enzyme-substrate pairings. In conclusion, our self-driving lab platform, equipped with a tailored optimization algorithm, offers a novel and superior alternative to traditional optimization methods. Moreover, the methodology for selecting the most efficient problem-specific optimization algorithm can be extended to self-driving lab platforms with broader applications.