TADA

TADA augments microbiome training datasets with taxonomy-aware synthetic samples to improve phenotype classification from high-dimensional, low-sample-size microbial profiles.

Key Features:

• Phylogenetic Augmentation: Leverages phylogenetic relationships among microbial taxa to guide generation of biologically relevant synthetic samples.
• Statistical Generative Model: Uses a statistical generative model to create realistic microbiome profiles for augmentation.
• Synthetic Sample Generation: Generates synthetic samples that preserve evolutionary connections and biological integrity.
• Addressing Data Imbalance: Augments under-represented classes to mitigate class imbalance and bias in microbiome datasets.
• Improved Machine Learning Robustness: Enhances robustness and accuracy of machine learning models applied to phenotype classification on high-dimensional, low-sample-size data.

Scientific Applications:

• Phenotype Classification: Improves predictive performance of models associating microbial compositions with specific host traits.
• Class Imbalance Scenarios: Applied where class imbalance and insufficient signal hinder traditional machine learning in microbiome studies.

Methodology:

Leverages phylogenetic relationships to generate synthetic microbiome samples via a statistical generative model and augments under-represented classes to balance training data.