MetaHD

MetaHD implements a multivariate meta-analysis model for high-dimensional metabolomics data that accounts for correlation between metabolites, within- and between-study variances, and missing values.

Key Features:

• Correlation modeling: Accommodates correlation between metabolites in the meta-analysis framework.
• Variance decomposition: Models both within-study and between-study variances.
• Missing data handling: Handles missing values and exploits borrowing strength between metabolites to improve estimates when data are missing at random.
• Data integration: Supports integration and collective analysis of individual-level metabolomics data from multiple studies as well as combination of summary estimates.

Scientific Applications:

• Metabolomics meta-analysis: Combines metabolomics data across multiple studies while maintaining statistical properties superior to existing approaches.
• Biomarker identification: Reduces misidentification of biomarkers by properly accounting for metabolite correlation and study variability.

Methodology:

MetaHD uses a multivariate meta-analysis model that accommodates correlation between metabolites, within- and between-study variances, and missing values; the model exploits borrowing strength between metabolites to improve estimation compared to univariate approaches.