Large-scale untargeted LC-MS metabolomics data correction using between-batch feature alignment and cluster-based...

Area of Science:

• Metabolomics
• Analytical Chemistry
• Bioinformatics

Background:

• Untargeted metabolomics using LC-MS offers broad metabolite coverage but suffers from batch variability.
• Measurement errors in signal intensity, mass accuracy, and retention times reduce data reproducibility.
• This variability hinders biological response detection and interpretation.

Purpose of the Study:

• To develop and implement robust procedures for correcting within- and between-batch variability in LC-MS metabolomics data.
• To enhance the quality and reliability of multi-batch untargeted metabolomics datasets.
• To improve the power for detecting biological signals and facilitate data interpretation.

Main Methods:

• Developed algorithms for systematic between-batch feature alignment and merging.
• Implemented orthogonal feature combination for improved cross-batch consistency.
• Utilized a cluster-based approach for within-batch drift correction accommodating multiple patterns.
• Created a heuristic criterion for selecting between-batch normalization strategies.

Main Results:

• Between-batch alignment increased feature detection by 15% and corrected 15% of misaligned features.
• Within-batch correction reduced the median coefficient of variation for quality control features from 20.5% to 15.1%.
• The developed algorithms are open-source and available as an R package ('batchCorr').

Conclusions:

• The developed procedures offer unbiased improvements in data quality for LC-MS metabolomics.
• These methods have significant implications for enhancing the accuracy of biological data analysis.
• The generic nature of these methods allows application to other data types with similar batch-related limitations.