AMAT

AMAT performs multivariate association testing of high-dimensional microbiome data by pooling microbial features into groups and applying distance correlation learning and adaptive testing within a generalized linear model framework to detect associations with health outcomes.

Key Features:

• High-Dimensional Data Handling: Pools individual microbial features into groups for multivariate association analysis to capture complex interactions within the microbiome.
• Feature Selection and Noise Reduction: Uses distance correlation learning for feature selection and to reduce noise accumulation in high-dimensional datasets.
• Adaptive Statistical Testing: Performs association testing within a flexible generalized linear model framework that accommodates various outcome types and distributional assumptions.
• Robustness and Power: Demonstrates, in simulation studies and real-world applications, preservation of the correct type I error rate while often improving statistical power relative to existing methods.

Scientific Applications:

• Microbiome association studies: Investigates associations between microbial communities and health outcomes, including gut microbiota-linked conditions such as inflammatory bowel disease, obesity, and diabetes.
• Biomarker and target discovery: Identifies potential microbial biomarkers or therapeutic targets within the microbiome through multivariate association analysis.

Methodology:

Pools microbial features into groups, applies distance correlation learning for feature selection and noise reduction, and conducts adaptive association testing within a generalized linear model framework.