Development of a Prediction Method of Cell Density in Autotrophic/Heterotrophic Microorganism Mixtures by Machine...

Area of Science:

• Biotechnology
• Machine Learning
• Spectroscopy

Background:

• Microflora are crucial for industrial value-added material production.
• Precise control of microflora cell density is essential for process stability.
• Existing methods for cell density evaluation can be complex or time-consuming.

Purpose of the Study:

• To develop a simple, AI-driven system for evaluating microflora cell density.
• To utilize absorbance spectra data for predicting the cell density of mixed microbial cultures.
• To enhance the industrial application of microflora through accurate and efficient cell density monitoring.

Main Methods:

• Constructed a machine learning model using absorbance spectra as features.
• Employed extremely randomized trees (ERT) for cell density prediction.
• Utilized Shapley Additive Explanations (SHAP) for model interpretability.

Main Results:

• Achieved high prediction accuracy (R²=0.8495 and R²=0.9232) for mixed cultures of Saccharomyces cerevisiae and Chlamydomonas reinhardtii.
• SHAP analysis identified key spectral features, including Soret and Q bands from chloroplasts, contributing to accurate predictions.
• Demonstrated the system's ability to differentiate and quantify cell densities in microbial mixtures.

Conclusions:

• The developed AI system provides a simple and effective method for evaluating microflora cell density.
• The system leverages spectral data and machine learning for robust predictions.
• This technology holds significant potential for industrial applications requiring precise microbial population control.