Synergistic enhancement of microalgal fatty acids via Batch Bayesian Optimization-driven combinatorial additive...

The industrial production of fatty acids by Schizochytrium sp. is constrained by low fatty acid content. While chemical regulation represents a promising non-genetic strategy to enhance lipid synthesis, its application is limited by the inefficient optimization of multi-additive combinations due to vast experimental spaces and complex interactions. To address this, the study developed a machine learning driven experimental design platform utilizing Batch Bayesian Optimization to efficiently navigate this high-dimensional parameter space. From an initial set of 13 candidates, four key additives were selected. Over six iterative rounds of optimization, the model's success rate for recommending effective combinations increased from 0.00 % to 77.50 %. Validation experiments confirmed 100.00 % reproducibility. This approach generated seven optimized regimens, including binary, ternary, and quaternary combinations, which enhanced the total fatty acids (TFA) content by approximately 50.00 %, achieving a final TFA content exceeding 50.00 % of dry cell weight in shake-flask cultures without significant growth inhibition, which represents the highest TFA content reported for Schizochytrium sp. under shake-flask conditions. This work establishes the power of machine learning driven optimization to unlock the synergistic potential of chemical additives, providing an efficient and scalable strategy for the sustainable microbial production of nutritionally valuable lipids.