GEEMiHC

GEEMiHC applies an adaptive generalized estimating equations (GEE)-based microbiome higher criticism to detect sparse microbial association signals in longitudinal microbiome data.

Key Features:

• Adaptive Microbiome Higher Criticism: Employs a higher criticism analysis tailored for microbiome data to detect sparse association signals in longitudinal studies.
• Generalized Estimating Equations (GEE) Framework: Utilizes GEE to model and account for correlations among repeated observations within individuals using various working correlation models.
• Robustness to Correlation Structures: Designed to maintain performance across diverse underlying correlation patterns in longitudinal data.
• Statistical Power and Type I Error Control: Demonstrates improved power to detect true associations while controlling type I error rates based on simulation experiments.
• Application to Real-World Longitudinal Data: Applied to longitudinal microbiome datasets to reveal associations such as those between the gut microbiome and Crohn's disease.
• Ranking of Significant Factors: Ranks factors associated with host phenotypes to aid identification of potential biomarkers.

Scientific Applications:

• Human health and disease: Identification of microbial associations with host phenotypes, including associations related to Crohn's disease.
• Longitudinal microbiome studies: Analysis of temporal changes and sparse signals in repeated-measures microbiome data.
• Dietary and environmental influence studies: Detection of subtle, infrequent microbial associations related to diet or environmental exposures over time.
• Biomarker prioritization: Ranking of associated factors to support downstream biomarker investigation.

Methodology:

Integrates multiple microbiome higher criticism analyses within a GEE framework and adaptively sets working correlation structures to remain robust across varying sparsity levels and phylogenetic relevance.