RAMClust: a novel feature clustering method enables spectral-matching-based annotation for metabolomics data

Area of Science:

• Metabolomics
• Analytical Chemistry
• Bioinformatics

Background:

• Metabolomic data analysis relies on feature detection (mass and retention time).
• A single metabolite can generate multiple mass signals, complicating analysis.
• Accurate metabolite identification is crucial but challenging.

Purpose of the Study:

• To develop a novel unsupervised method for grouping mass spectrometry (MS) signals into spectra.
• To improve metabolite annotation by integrating MS and indiscriminant MS/MS (idMS/MS) data.
• To reduce quantitative variation and false positive annotations in metabolomic studies.

Main Methods:

• Implemented an unsupervised feature grouping method to cluster MS signals into spectra.
• Incorporated idMS/MS data implicitly by performing feature detection on both MS and idMS/MS data.
• Determined feature-feature relationships simultaneously from MS and idMS/MS data.

Main Results:

• The novel method effectively groups MS signals into spectra without predicting in-source phenomena.
• Simultaneous analysis of MS and idMS/MS data facilitates metabolite identification.
• The approach reduces quantitative analytical variation and minimizes false positive annotations.

Conclusions:

• RAMClustR is a versatile R package for grouping features from various chromatographic-MS platforms.
• This method enhances metabolite identification accuracy and data reliability in metabolomics.
• The tool offers a solution for a fundamental bottleneck in MS-based metabolomic data analysis.