Filtering procedures for untargeted LC-MS metabolomics data

Area of Science:

• Metabolomics
• Biochemistry
• Bioinformatics

Background:

• Untargeted metabolomics data often contains numerous uninformative features.
• These features can hinder crucial analyses like biomarker discovery and metabolic pathway analysis.
• A need exists for adaptable methods to pre-process metabolomics data effectively.

Purpose of the Study:

• To propose and evaluate a novel data-adaptive pipeline for filtering untargeted metabolomics data.
• To enhance the quality of metabolomics datasets for downstream biological investigations.
• To provide a versatile tool for handling complex biospecimen data.

Main Methods:

• Development of a data-adaptive filtering pipeline for liquid chromatography-mass spectrometry (LC-MS) data.
• Incorporation of novel filtering criteria including blank samples, missing value proportions, and intra-class correlation coefficients.
• Utilized human blood and public LC-MS datasets for validation.

Main Results:

• The data-adaptive filtering method demonstrated superior performance compared to traditional threshold-based approaches.
• Effectively removed noisy features while preserving high-quality, biologically relevant information.
• R code for the pipeline is publicly available.

Conclusions:

• The proposed data-adaptive pipeline offers an intuitive and effective solution for filtering uninformative features in untargeted metabolomics.
• This method is particularly valuable for analyzing complex biological matrices.
• Enhances the reliability of biological phenomenon interrogation from metabolomics data.