Optimization of Liposome Production via Microfluidic Method: A Comparative Study of Design of Experiments Approaches

This study presents an optimized approach for liposome production using the microfluidic method by integrating Design of Experiments (DoE) and machine learning. Three DoE methodologies —Box-Behnken Design (BBD), Central Composite Design (CCD), and Full Factorial Design—were systematically compared to identify the most efficient strategy for process optimization while minimizing the number of experimental runs. Process modeling was performed using the Gradient Boosting Regressor algorithm, with model performance assessed based on R², MAE, and RMSE metrics. The findings demonstrated that the CCD approach achieved the highest predictive accuracy for liposome size (R² = ۰.۹۸۷۰) with a reduced number of experiments. Conversely, the full factorial design yielded comparable accuracy but proved inefficient in terms of time and resource allocation due to the extensive number of required experiments. The BBD method was deemed unsuitable due to its lower predictive accuracy. This study underscores the potential of leveraging DoE in conjunction with machine learning to enhance liposome production efficiency and reduce experimental costs.

کلیدواژه ها:

Design of Experiments ، Liposome Production ، Microfluidic Method ، Optimization ، Machine Learning