SingleFrag: a deep learning tool for MS/MS fragment and spectral prediction and metabolite annotation

Area of Science:

• Analytical Chemistry
• Computational Chemistry
• Biochemistry

Background:

• Metabolite and small molecule identification relies on matching experimental tandem mass spectrometry (MS/MS) spectra with reference libraries.
• The scarcity of comprehensive spectral libraries impedes accurate compound identification.
• Accurate in silico fragmentation prediction is crucial for generating MS/MS spectra of compounds lacking empirical data.

Purpose of the Study:

• To develop and validate a novel deep learning tool, SingleFrag, for predicting MS/MS spectra.
• To address the limitations of current in silico fragmentation tools by predicting individual fragments.
• To enhance the annotation of unknown MS/MS spectra.

Main Methods:

• Development of SingleFrag, a deep learning model predicting individual molecular fragments.
• Evaluation of SingleFrag's performance against state-of-the-art in silico fragmentation tools.
• Application of SingleFrag for annotating unknown MS/MS spectra of compounds in human samples.

Main Results:

• SingleFrag demonstrates superior performance compared to existing state-of-the-art in silico fragmentation tools.
• The tool accurately predicts individual fragments, contributing to more reliable spectral prediction.
• Successfully annotated three previously unidentified compounds commonly found in human samples.

Conclusions:

• SingleFrag offers a powerful and accurate method for in silico MS/MS spectral prediction.
• This tool can significantly aid in the annotation of unknown MS/MS spectra, particularly for compounds without existing reference spectra.
• SingleFrag advances metabolite identification capabilities, with potential applications in various biological and chemical analyses.