With the increased use of metabolomics and global efforts towards science transparency, the amount of publicly available metabolomics data deposited in public repositories such as Metabolomics Workbench, MetaboLights and OmicsDI has grown tremendously.
Integrating datasets collected on the same or similar conditions from independent studies, a process called meta-analysis, can help address the reproducibility issues due to low sample size, sample heterogeneity, and identify robust patterns and biomarkers. MetaboAnalyst implements two relatively straightforward approaches to support meta-analysis in the Functional Meta-analysis module.
Feature level meta-analysis
This is a low-level integration where features are merged directly for enrichment analysis. Such practice is usually restricted to the data from the same studies. For instance, in LC-HRMS metabolomics, a common practice is to collect spectra from the same samples using both positive and negative ion modes. The resulting peaks can be directly merged for downstream pathway activity analysis.
Pathway level meta-analysis
In this case, LC-MS peak tables should be collected under similar experimental and analytical conditions (i.e. similar metabolome coverage). Each datasets are subject to pathway enrichment analysis, and the resulting pathway-level p values are then integrated based on fixed or random effects models. More details are here.