MetaFX: feature extraction from whole-genome metagenomic sequencing data

Area of Science:

• Metagenomics
• Bioinformatics
• Computational Biology

Background:

• Microbial communities significantly impact host health, but direct links between specific microbes and host states are often unclear.
• Current reference-based tools are limited by database inclusions, while reference-free methods are resource-intensive and yield less interpretable results.
• Accurate analysis of complex microbial communities is crucial for understanding host-environment interactions.

Purpose of the Study:

• To introduce MetaFX, an open-source library for feature extraction from whole-genome metagenomic data.
• To enable classification of sample groups based on distinct microbial community genomic features.
• To provide a tool for biological insights through visualization and annotation of genomic features.

Main Methods:

• MetaFX utilizes statistical k-mer analysis and de Bruijn graph partitioning to construct genomic features from metagenomic data.
• These features are then employed in machine learning models for sample classification.
• The library facilitates comparison of samples grouped by metadata, such as disease or treatment status.

Main Results:

• MetaFX successfully built classification models for human gut samples, achieving up to 17% higher disease prediction accuracy for inflammatory bowel disease.
• The toolkit demonstrated improved classification results compared to traditional taxonomic analysis, with an average improvement of 9±10%.
• Extracted features can be visualized on de Bruijn graphs, aiding biological interpretation.

Conclusions:

• MetaFX offers an efficient and effective approach for feature extraction and classification in metagenomic datasets.
• The library enhances the accuracy and interpretability of microbial community analysis.
• MetaFX provides a valuable resource for researchers studying host-microbe interactions and disease states.