MetaboVariation

MetaboVariation models repeated metabolite measurements at the individual level using a Bayesian generalized linear model to identify intra-individual metabolic variations indicative of early metabolic dysfunction.

Key Features:

• Bayesian Generalized Linear Model: Uses a Bayesian generalized linear model to capture intra-individual variation in metabolite levels across multiple time points.
• Posterior Predictive Distribution: Generates posterior predictive distributions and flags observations that fall outside the 95% highest posterior density prediction interval at specific time points.
• Explanatory Variables Integration: Models repeated metabolite levels as a function of explanatory variables while accounting for intra-individual variation.

Scientific Applications:

• Individual-level biomarker identification: Identifies individuals with significant intra-individual variations in metabolites to inform detection of early metabolic dysfunction.
• Longitudinal dataset analysis: Demonstrated on 20 metabolites measured quarterly in 164 individuals, flagging 28% of individuals with variations in three or more metabolites.

Methodology:

Models repeated measurements on individuals using a Bayesian generalized linear model, generates posterior predictive distributions, and flags observations outside the 95% highest posterior density prediction interval while incorporating explanatory variables to account for intra-individual variation.