Applying machine learning to classify table olives using bacterial metataxonomic data.

Area of Science:

• Microbiology
• Bioinformatics
• Food Science

Background:

• Metataxonomic analysis is crucial for understanding microbial communities in fermented foods.
• Bioinformatics and machine learning (ML) offer advanced tools for analyzing complex microbial data.
• Tree-based ML algorithms provide interpretable insights into metataxonomic datasets.

Purpose of the Study:

• To compare three tree-based ML algorithms for analyzing 16S rRNA bacterial profiles from table olives.
• To assess the effectiveness of ML in classifying bacterial communities based on food characteristics.
• To identify the most accurate ML model for this application.

Main Methods:

• Analysis of 442 table olive samples using 16S rRNA sequencing.
• Application and comparison of Classification and Regression Tree, Random Forest (RF), and Extreme Gradient Boosting algorithms.
• Evaluation of classification accuracy based on processing type, cultivar, country of origin, and isolation matrix.

Main Results:

• ML techniques effectively classified bacterial profiles according to various food attributes.
• The Random Forest (RF) model demonstrated the highest accuracy, reaching up to 97%.
• RF achieved a kappa coefficient above 0.8 for most classification categories, indicating strong performance.

Conclusions:

• Tree-based ML, particularly RF, is a powerful tool for metataxonomic analysis of table olive microbial communities.
• This approach can significantly enhance traceability, authenticity, and quality control in the food industry.
• The findings have potential applications beyond table olives to other fermented food products.