Machine learning strategy to boost 1-deoxynojirimycin production through fermentation by Bacillus amyloliquefaciens

Area of Science:

• Biotechnology
• Bioprocess Engineering
• Machine Learning Applications

Background:

• 1-deoxynojirimycin (1-DNJ) is a valuable natural alkaloid with pharmaceutical and nutraceutical uses.
• Microbial biosynthesis presents a sustainable production method, but optimization is challenging due to complex fermentation parameters.

Purpose of the Study:

• To enhance 1-deoxynojirimycin (1-DNJ) production by Bacillus amyloliquefaciens.
• To apply a combination of single-factor experiments and artificial neural network (ANN) models for process optimization.

Main Methods:

• Utilized LC-MS/MS for 1-DNJ confirmation and quantification.
• Employed single-factor experiments for initial medium optimization.
• Developed and trained an ANN model (rprop+ algorithm, 6 hidden neurons) on experimental data.

Main Results:

• Single-factor optimization yielded 1064 ± 4.64 mg/L 1-DNJ with 82.55% alpha-glucosidase inhibition.
• The ANN model predicted optimized conditions achieving 1179 ± 8.44 mg/L 1-DNJ and 87.47% inhibition.
• Sensitivity analysis identified inoculum volume and fermentation temperature as key variables.

Conclusions:

• The integrated machine learning framework effectively optimizes 1-DNJ fermentation.
• Identified critical fermentation parameters provide a basis for future research and scale-up.