MIRTH

MIRTH imputes unmeasured metabolite abundances across heterogeneous metabolomics datasets to recover missing metabolic features and enable integrated analyses of metabolite covariation.

Key Features:

• Imputation of Unmeasured Metabolites: Imputes unmeasured metabolite abundances by leveraging latent covariation information embedded within measured metabolites.
• Rank-Transformation and Harmonization: Applies rank-transformation techniques and harmonization strategies to align heterogeneous datasets and correct systematic biases before imputation.
• Joint Modeling Across Datasets: Jointly models metabolite covariation across multiple independently-profiled datasets to recover masked metabolite abundances within and across studies.

Scientific Applications:

• Hypothesis Generation: Reveals imputed metabolites that enable formulation of hypotheses about metabolic pathways and interactions that are obscured by incomplete coverage.
• Workflow Simplification: Reduces dependence on extensive experimental replication and platform-specific coverage by providing imputed metabolite profiles across platforms.

Methodology:

MIRTH identifies patterns of covariation among measured metabolites, applies rank-transformation and harmonization across datasets, and jointly models covariation to infer abundances of unmeasured metabolites.