Fragment Distiller: Data Driven Discovery of Fragmentation Patterns in Tandem Mass Spectrometry

Tandem Mass Spectrometry (MS/MS) fragmentation patterns serve as a reproducible fingerprint for small molecules in experimental data. These MS/MS spectra yield insight into the chemical structure of the original compound. Specifically, certain conserved substructures yield reproducible patterns of fragmentation. Here, we tackle the challenge of identifying consistent fragmentation patterns in MS/MS data in a data driven fashion. We introduce Fragment Distiller, a computational tool that leverages publicly available MS/MS spectral libraries to discover consistent fragmentation patterns across any user-defined subset of molecules. In this manuscript, we establish the applicability of Fragment Distiller by extracting fragmentation fingerprints of two compound classes, carnitines and bile acids, codifying these patterns as re-usable knowledge, and applying this knowledge to discover dozens of new metabolites in public mass spectrometry data on the repository scale. We anticipate Fragment Distiller will occupy a key role in a broader mass spectrometry data ecosystem, as this tool arms the community with the ability to extract new knowledge from existing MS/MS library resources and apply this knowledge to accelerate compound discovery in the metabolomics dark matter. We expect the application of Fragment Distiller will increase the utility and incentive for deposition of new MS/MS library spectra into public repositories by creating a positive feedback cycle of discovery that further enhances Fragment Distiller’s capabilities.