Develop a hybrid machine learning model for promoting microbe biomass production.

Area of Science:

• Biotechnology
• Mycology
• Machine Learning Applications

Background:

• Optimizing microbial biomass production, specifically mycelia yield for Antrodia cinnamomea, is complex due to numerous cultivation factors.
• Traditional methods for identifying optimal conditions are time-consuming and labor-intensive.

Purpose of the Study:

• To develop and apply a hybrid machine learning approach, ANFIS-NM, for identifying critical factors influencing Antrodia cinnamomea cultivation.
• To optimize the cultivation conditions for enhanced mycelia yield using the proposed ANFIS-NM model.

Main Methods:

• A 32 fractional factorial design with seven factors was employed to screen potential cultivation variables.
• The Adaptive Neuro-Fuzzy Inference System with Novelty Mining (ANFIS-NM) was utilized to analyze experimental data and identify key factors.
• A central composite design (three factors, five levels) was subsequently used for fine-tuning and determining optimal yield conditions.

Main Results:

• The ANFIS-NM approach successfully pinpointed glucose, potato dextrose broth, and agar as the most potent factors influencing mycelia yield.
• Implementation of the optimized conditions led to a significant increase in Antrodia cinnamomea mycelia yield.
• The interpretability of the ANFIS rules provided visual insights into the cultivation condition relationships.

Conclusions:

• The hybrid ANFIS-NM approach offers an efficient method for optimizing microbial cultivation, significantly reducing laboratory time.
• This study establishes a new benchmark for biomass production, aligning with Sustainable Development Goals 7 and 12.
• The ANFIS-NM model demonstrates substantial potential for accelerating research and development in fungal biotechnology.