EXPERT: transfer learning-enabled context-aware microbial community classification.

Area of Science:

• Microbiology
• Bioinformatics
• Computational Biology

Background:

• Microbial community classification is crucial for understanding microbial ecology and its role in health and disease.
• Existing classification models often struggle with a speed-accuracy trade-off and lack adaptability to diverse or understudied contexts.
• Customization is challenging, limiting the application of microbial classification in various research scenarios.

Purpose of the Study:

• To introduce EXPERT, a novel transfer learning-based model for microbial community classification.
• To demonstrate the adaptability, high efficiency, and accuracy of EXPERT across multiple contexts.
• To showcase the utility of transfer learning in microbial classification for both disease-specific and longitudinal microbiome studies.

Main Methods:

• Development of EXPERT utilizing transfer learning principles.
• Application of EXPERT to classify microbial communities in diverse datasets, including those from multiple diseases with limited samples.
• Validation of EXPERT's performance in predicting gut microbiome changes during colorectal cancer progression.

Main Results:

• EXPERT demonstrated high efficiency and accuracy in microbial community classification.
• The model proved adaptable to various contexts, including limited-sample disease datasets.
• EXPERT successfully predicted gut microbiome alterations across colorectal cancer stages.

Conclusions:

• EXPERT provides accurate and context-aware microbial community classification.
• Transfer learning enhances the adaptability and performance of microbial classification models.
• EXPERT facilitates novel microbial knowledge discovery and supports customized microbiome research.