GutBalance

GutBalance applies Distal Discriminative Balance Analysis (DBA) to species-level shotgun metagenomic abundances from GMrepo to derive compositional balances as biomarkers and to build balance-based disease prediction models.

Key Features:

• Distal Discriminative Balance Analysis (DBA): Selects distal balances from pairs and trios of bacterial species to manage high-dimensional compositional microbiome data for classification tasks.
• Model Repository: Applies DBA to species-level abundances across disease phenotypes in GMrepo to construct a balance-based model repository for predicting disease risk in new samples.
• Balance-Disease Associations: Emphasizes balance-disease associations in which each balance represents a bacteria ratio correlated positively or negatively with a disease state and can serve as a biomarker.
• Human Gut Balance-Disease Association Database (GBDAD): Provides a structured database of identified balance-disease associations to facilitate interpretation of biomarker–disease relationships.
• Validation and Hypothesis Generation: Links balance-disease associations to microbe-disease associations from MicroPhenoDB to validate species-disease associations inferred from shotgun metagenomic datasets and support hypothesis generation.
• Compositional Transformation Handling: Addresses compositional challenges and compositional transformation during supervised learning to improve the reliability of disease classifiers.

Scientific Applications:

• Disease Risk Prediction: Predicts disease risk from human gut microbiome species-level abundances using balance-based models.
• Biomarker Discovery: Identifies balance-based biomarkers (bacteria ratios) associated with gastrointestinal diseases.
• Hypothesis Generation and Validation: Generates and validates hypotheses about microbe–disease interactions by linking balance-disease associations to MicroPhenoDB and existing species-disease evidence.
• Gastrointestinal Microbiome Research: Supports studies of gastrointestinal diseases using shotgun metagenomic species-level abundance data.

Methodology:

Apply DBA to species-level shotgun metagenomic abundances from GMrepo to select distal balances from pairs and trios, construct a balance-based model repository and GBDAD, and link identified balances to MicroPhenoDB for external validation of species-disease associations.