DisBalance

DisBalance constructs balance-based disease prediction models and discovers microbial taxonomic biomarkers from compositional microbiome and shotgun metagenomic data.

Key Features:

• DBA-distal: Employs Distal Discriminative Balances Analysis (DBA-distal) to select distal discriminative balances that prioritize interpretability, runtime efficiency, and classification accuracy.
• Integrated workflow: Integrates workflows for disease-model construction, risk prediction, and taxonomic biomarker discovery in microbiome-based binary classifications.
• Model building and prediction: Supports de novo model-building and rapid disease risk prediction using microbial taxonomic abundances.
• Biomarker mining strategies: Implements multiple model-driven and knowledge-driven strategies to mine microbial biomarkers.
• Validation and application: Validated by independent testing on seven microbiome datasets and demonstrated with a complete analysis of a shotgun metagenomic Ulcerative Colitis dataset.

Scientific Applications:

• Microbiome-based disease prediction: Binary classification and risk prediction of human diseases using balance-based models.
• Taxonomic biomarker discovery: Identification of microbial taxonomic biomarkers associated with disease via model-driven and knowledge-driven mining.
• Ulcerative Colitis analysis: Analysis and interpretation of shotgun metagenomic data for Ulcerative Colitis studies.

Methodology:

Uses Distal Discriminative Balances Analysis (DBA-distal) to select distal discriminative balances, performs de novo model-building and risk prediction from microbial taxonomic abundances, applies model-driven and knowledge-driven biomarker mining strategies, and validates results on seven independent microbiome datasets including a shotgun metagenomic Ulcerative Colitis dataset.