Hybrid feature detection and information accumulation using high-resolution LC-MS metabolomics data

Area of Science:

• Metabolomics
• Computational Biology
• Analytical Chemistry

Background:

• Feature detection is crucial for liquid chromatography-mass spectrometry (LC-MS) metabolomics data preprocessing.
• Current methods rely solely on noise filters and peak shape models, limiting sensitivity for low-concentration metabolites.
• Leveraging metabolite databases and historical data is challenging due to high noise levels and potential false positives in LC-MS data.

Purpose of the Study:

• To develop a computational approach that enhances feature detection sensitivity in LC-MS metabolomics.
• To reduce false positives in feature detection by incorporating novel algorithms.
• To enable accumulation of metabolite concentration variation data across samples for future rare or uncommon feature identification.

Main Methods:

• A hybrid procedure combining untargeted and targeted peak detection was employed.
• New algorithms for nonparametric local peak detection and filtering were designed to minimize false positives.
• The approach was implemented as part of the R package apLCMS.

Main Results:

• The computational approach demonstrated improved feature detection sensitivity.
• The method effectively reduced the occurrence of false positives in LC-MS data.
• The study successfully showcased the value of the approach in a proof-of-concept analysis.

Conclusions:

• The developed computational approach significantly boosts feature detection sensitivity in LC-MS metabolomics.
• The method provides a robust strategy for reducing false positives, enhancing metabolite identification accuracy.
• This approach facilitates the accumulation of valuable concentration variation data for future discoveries in metabolomics research.