BUDDY

BUDDY performs de novo molecular formula annotation from tandem mass spectrometry (MS/MS) data to improve metabolite identification in MS-based metabolomics.

Key Features:

• Bottom-Up Tandem MS Interrogation: Employs bottom-up MS/MS interrogation that prioritizes candidate molecular formulae explainable by observed MS/MS fragments.
• Machine-Learned Ranking and FDR Estimation: Uses machine learning to rank candidate molecular formulae and provides false discovery rate (FDR) estimation for annotation confidence.
• Reduction in Formula Candidate Space: Reduces formula candidate space by an average of 42.8% compared to exhaustive mathematical enumeration.
• Systematic Benchmarking: Benchmarked on reference MS/MS libraries and real metabolomics datasets to evaluate annotation accuracy.
• Novel Molecular Formula Annotation: Annotated over 5,000 novel molecular formulae from 155,321 recurrent unidentified spectra that were absent from existing chemical databases.
• Global Optimization and Peak Interrelationships: Integrates global optimization with bottom-up MS/MS interrogation to refine annotations and reveal interrelationships between peaks.
• Application in Fatty Acid Amide Annotation: Applied to human fecal data to systematically annotate 37 fatty acid amide molecules.

Scientific Applications:

• De novo metabolite annotation: Annotation of molecular formulae for unidentified MS features using MS/MS evidence.
• Novel metabolite discovery: Identification of molecular formulae absent from chemical databases to enable discovery of new biochemical entities.
• Dataset-level interpretation: Refinement of annotations and inference of peak interrelationships via global optimization for comprehensive metabolomic dataset analysis.
• Targeted biochemical class annotation: Systematic annotation of fatty acid amides in human fecal metabolomics studies.

Methodology:

Bottom-up tandem MS (MS/MS) interrogation for de novo formula annotation; machine-learned ranking with false discovery rate (FDR) estimation; global optimization integrated with bottom-up MS/MS interrogation; comparison against exhaustive mathematical enumeration; systematic benchmarking on reference MS/MS libraries and real metabolomics datasets.