GRaMM

GRaMM identifies associations between the metabolome and microbiome by integrating linear regression, the maximum information coefficient (MIC), metabolic confounding effect elimination (MCEE), and centered log-ratio (CLR) transformation to capture linear and nonlinear correlations while addressing confounding and compositional biases.

Key Features:

• Integration of complementary methods: Combines linear regression and maximum information coefficient (MIC) for linear and nonlinear correlation detection together with metabolic confounding effect elimination (MCEE) and centered log-ratio (CLR) transformation to manage confounders and compositional data.
• Sequential computational steps: Implements data preprocessing, linear versus nonlinear type identification, confounder correction and correlation detection, and p-value correction as explicit analysis steps.
• Handling of omics data characteristics: Explicitly addresses metabolomic and microbiome data-specific challenges including confounding effects and compositional biases.
• Performance evaluation: Evaluated on multiple simulated and real-world datasets with reported high accuracy, sensitivity, specificity, and low false positive rate, and assessment of the impact of preprocessing and confounder adjustment.
• Comparative advantage: Presented as a strategy specifically designed for intercorrelation analysis between metabolites and microbes, with improvements in accuracy and robustness over existing methods.

Scientific Applications:

• Microbe–metabolite association discovery: Identification of statistically significant associations between microbes and metabolites in multi-omics studies.
• Physiological and disease mechanism studies: Investigation of metabolome–microbiome intercorrelations to explore physiological roles and disease mechanisms.
• Therapeutic strategy development: Support for studies that aim to derive insights relevant to novel therapeutic strategies based on microbe–metabolite interactions.

Methodology:

Performs data preprocessing, identifies linear versus nonlinear associations using linear regression and maximum information coefficient (MIC), applies metabolic confounding effect elimination (MCEE) and centered log-ratio (CLR) transformation, and conducts p-value correction.