Active learning sequential surrogate-based optimization (AL-SSBO) versus response surface methodology in sunflower oil diacylglycerol synthesis: Performance and insight into AL-SSBO.

Diacylglycerol (DAG) are functional lipids with benefits including weight reduction and improved blood lipid profiles. However, enzymatic synthesis optimization is constrained by the rigid design of conventional response surface methodology (RSM). This study experimentally implemented and evaluated an Active-Learning Sequential Surrogate-Based Optimization workflow for sunflower oil DAG synthesis. AL-SSBO sequentially incorporated replicate measurements under recommended conditions and achieved a DAG content of 53.77% after five predefined optimization rounds. The AL-SSBO yielded a prediction error of only 0.10 percentage points, compared with 1.43 for RSM. Bootstrap-based inference supports the observed performance advantage of AL-SSBO over RSM. Optimal conditions were 11.93 h, 68.56 °C, 8.93 wt% lipase loading, and an oil-to-glycerol weight ratio of 11.71. Analysis of exploration-exploitation behavior and spatial sampling showed that the recommended conditions became increasingly concentrated near a high-DAG region, consistent with retrospective stabilization within the investigated domain.