Computational Analyses of Spectral Trees from Electrospray Multi-Stage Mass Spectrometry to Aid Metabolite Identification

Mass spectrometry coupled with chromatography has become the major technical platform in metabolomics. Aided by peak detection algorithms, the detected signals are characterized by mass-over-charge ratio (m/z) and retention time. Chemical identities often remain elusive for the majority of the signals. Multi-stage mass spectrometry based on electrospray ionization (ESI) allows collision-induced dissociation (CID) fragmentation of selected precursor ions. These fragment ions can assist in structural inference for metabolites of low molecular weight. Computational investigations of fragmentation spectra have increasingly received attention in metabolomics and various public databases house such data. We have developed an R package “iontree” that can capture, store and analyze MS(2) and MS(3) mass spectral data from high throughput metabolomics experiments. The package includes functions for ion tree construction, an algorithm (distMS2) for MS(2) spectral comparison, and tools for building platform-independent ion tree (MS(2)/MS(3)) libraries. We have demonstrated the utilization of the package for the systematic analysis and annotation of fragmentation spectra collected in various metabolomics platforms, including direct infusion mass spectrometry, and liquid chromatography coupled with either low resolution or high resolution mass spectrometry. Assisted by the developed computational tools, we have demonstrated that spectral trees can provide informative evidence complementary to retention time and accurate mass to aid with annotating unknown peaks. These experimental spectral trees once subjected to a quality control process, can be used for querying public MS(2) databases or de novo interpretation. The putatively annotated spectral trees can be readily incorporated into reference libraries for routine identification of metabolites.