Related Experiment Video
Updated: Aug 9, 2025
08:27
Large-Scale Multi-Omics Genome-Wide Association Studies Mo-GWAS: Guidelines for Sample Preparation and Normalization
Published on: July 27, 2021
3.8K
Regularized adversarial learning for normalization of multi-batch untargeted metabolomics data
Andrei Dmitrenko1,2, Michelle Reid1, Nicola Zamboni1,3
1ETH Zürich, Institute of Molecular Systems Biology, Zürich 8093, Switzerland.
Bioinformatics (Oxford, England)
|February 24, 2023
Summary
We developed Regularized Adversarial Learning Preserving Similarity (RALPS) to normalize untargeted metabolomics data. This method effectively reduces batch effects while preserving biological information, improving multi-batch study analysis.
Area of Science:
- Metabolomics
- Mass Spectrometry
- Bioinformatics
Background:
- Untargeted metabolomics enables unbiased analysis of complex biological samples.
- High-resolution mass spectrometry profiles thousands of molecules, but inter-batch differences pose challenges.
- Batch effects hinder multi-batch untargeted metabolomics studies and inter-laboratory comparisons.
Purpose of the Study:
- To introduce a novel method for normalizing multi-batch untargeted metabolomics data.
- To address the unresolved problem of inter-batch differences in metabolomics.
- To improve the reliability and comparability of untargeted metabolomics experiments.
Main Methods:
- Developed Regularized Adversarial Learning Preserving Similarity (RALPS).
- Utilized deep adversarial learning with a three-term loss function.
- Ensured mitigation of batch effects while preserving biological identity, spectral properties, and coefficients of variation.
Main Results:
- RALPS demonstrated superior performance compared to six state-of-the-art batch correction methods on two large metabolomics datasets.
- The method effectively mitigates batch effects in untargeted metabolomics data.
- RALPS shows scalability, robustness, handles missing values, and accommodates diverse experimental designs.
Conclusions:
- RALPS offers a robust solution for normalizing multi-batch untargeted metabolomics data.
- The method preserves crucial biological and spectral information, enhancing data integrity.
- RALPS facilitates more reliable analysis and intercomparison of metabolomics experiments.
