Optimizing agarase production from Microbulbifer sp. using response surface methodology and machine learning models.

Area of Science:

• Enzymology and biotechnology
• Microbial fermentation
• Bioinformatics

Background:

• Agarase enzymes are vital in diverse industries including food, cosmetics, and medicine.
• Current agarase production methods suffer from low yields, high costs, and inconsistent activity.
• There is a significant need for optimized microbial sources for efficient industrial-scale agarase production.

Purpose of the Study:

• To optimize extracellular agarase production from a microbial source, specifically *Microbulbifer* sp.
• To investigate and identify optimal growth conditions for enhanced agarase yield and activity.
• To explore machine learning algorithms for accurate prediction of agarase activity.

Main Methods:

• Qualitative-quantitative analysis of microbial growth conditions.
• Response Surface Methodology (RSM) to optimize agar concentration, pH, temperature, and incubation time.
• Application and evaluation of machine learning algorithms, including Radial Basis Function neural network, for predictive modeling.

Main Results:

• Optimized conditions determined as 0.3% agar, pH 7, 25°C, and 36-hour incubation, yielding 317.97 μmol min⁻¹ agarase activity.
• High statistical significance confirmed by F-value (44.75) and R-squared (0.9827) for experimental validation.
• Radial Basis Function neural network demonstrated superior predictive performance with R-squared of 0.989 and MSE of 0.44.

Conclusions:

• Optimized production parameters significantly enhance agarase scalability and efficiency.
• Machine learning models provide robust and accurate predictions for agarase activity.
• Findings support industrial-scale bioreactor operations with potential for real-time adjustments.