mzrtsim

mzrtsim simulates and evaluates batch effects in GC-MS and LC-MS metabolomics peak intensity data to support selection and assessment of batch correction methods.

Key Features:

• Simulation-Based Approach: Performs in silico simulations to model batch effects on metabolomics peak intensity data using statistical properties from existing datasets.
• Extensive Evaluation Framework: Evaluates batch correction methods across 252,000 batch corrections applied to 14,000 simulated datasets.
• Focus on Statistical Properties: Parameterizes simulations based on statistical properties specific to metabolomics data, reflecting differences from genomics data.
• Impact of Log Transformations: Assesses how log transformations affect batch correction performance, including effects on false positive and true positive rates.
• Recommendation for Preliminary Simulations: Recommends conducting preliminary simulations using real data distributions to inform selection of appropriate batch correction methods for a given dataset.

Scientific Applications:

• Validation of batch correction methods: Enables systematic validation and benchmarking of batch correction algorithms on simulated metabolomics datasets.
• Assessment of preprocessing strategies: Allows evaluation of preprocessing choices such as log transformation for their impact on correction performance.
• Experimental design for batch effect mitigation: Informs design of preliminary experiments and simulation studies to choose appropriate correction strategies for specific GC-/LC-MS datasets.

Methodology:

Uses in silico simulations of batch effects on metabolite peak intensities parameterized from statistical properties of existing metabolomics datasets, with large-scale evaluation comprising 252,000 batch corrections across 14,000 simulated datasets and analysis of the effects of log transformations on method performance.