Prediction of ethanol fermentation under stressed conditions using yeast morphological data

Area of Science:

• Biotechnology
• Microbiology
• Artificial Intelligence

Background:

• Yeast fermentation is crucial for biocommodity production but is sensitive to high-sugar stress.
• High-sugar conditions alter yeast intracellular states and cell morphology.

Purpose of the Study:

• To develop artificial intelligence (AI) models for predicting ethanol yields in yeast fermentation under high-sugar stress.
• To utilize cell morphological data for accurate yield prediction.

Main Methods:

• Extraction of high-dimensional morphological data from phase contrast images.
• Supervised machine learning, specifically neural network algorithms, for yield prediction.
• Analysis of morphological changes under varying glucose concentrations and fermentation phases.

Main Results:

• A neural network model achieved a high prediction accuracy (R² = 0.95).
• The model could predict ethanol yields up to 60 minutes in advance.
• Morphological data from low-glucose conditions were insufficient for high-glucose stress prediction.
• Cell morphology under high glucose resembled cells under high osmotic pressure.

Conclusions:

• Cell morphology analysis provides insights into yeast intracellular physiological states under stress.
• AI models based on morphology can effectively predict ethanol yields in stressed yeast cultures.
• Monitoring yeast morphology can improve the management and stable production of biocommodities.