Multispectrum ModiFinder Site Localization Performance

Tandem mass spectrometry (MS/MS) is a powerful technique
for structural
identification of small molecules, yet a significant portion of MS/MS
spectra from untargeted experiments remain unidentifiable through
spectrum library matching. ModiFinder, a computational tool, tackles
this issue by predicting the site of chemical modifications on known
analogs of the unidentified compounds using MS/MS data. However, ModiFinder’s
performance is limited by insufficient peak data and fragmentation
annotation ambiguities. In this study, we investigate how incorporating
MS/MS spectra from multiple collision energies and mass spectrometry
adducts can enhance ModiFinder’s localization accuracy. Using
a data set from Agilent Technologies comprising 2150 data-rich compounds
(five times larger than previously available data sets), we evaluated
the impact of complementary spectral information. Our results show
that combining spectra from different adducts and collision energies
expands ModiFinder’s localization abilities to more compounds
and improves the overall performance.