Multiple-Disease Detection and Classification across Cohorts via Microbiome Search

Area of Science:

• Microbiology
• Bioinformatics
• Computational Biology

Background:

• Microbiome-based disease classification often relies on specific models or markers, which are not available for many conditions.
• Existing methods face challenges with data heterogeneity and contamination in microbiome datasets.

Purpose of the Study:

• To introduce a novel search-based strategy for detecting and classifying diseases using microbiome data.
• To provide an alternative approach when disease-specific models or markers are absent.

Main Methods:

• The strategy identifies diseased samples by detecting their novelty as outliers against a healthy subject database.
• Diseased samples are then compared to databases of samples from known patients for classification.
• The approach utilizes 16S rRNA gene amplicon data.

Main Results:

• The search-based strategy demonstrated superior precision, sensitivity, and speed compared to traditional model-based approaches.
• The method proved robust against platform heterogeneity and contamination in sequencing data.
• It successfully identified microbiome states associated with disease across different cohorts and platforms.

Conclusions:

• The search-based strategy is a promising tool for microbiome big-data-based diagnosis.
• This approach facilitates disease detection and classification even with diverse datasets and potential contamination.
• It offers a robust alternative for identifying disease-associated microbiome signatures.