MetFusion: integration of compound identification strategies

Abstract

Mass spectrometry (MS) is an important analytical technique for the detection and identification of small compounds. The main bottleneck in the interpretation of metabolite profiling or screening experiments is the identification of unknown compounds from tandem mass spectra. Spectral libraries for tandem MS, such as MassBank or NIST, contain reference spectra for many compounds, but their limited chemical coverage reduces the chance for a correct and reliable identification of unknown spectra outside the database domain. On the other hand, compound databases like PubChem or ChemSpider have a much larger coverage of the chemical space, but they cannot be queried with spectral information directly. Recently, computational mass spectrometry methods and in silico fragmentation prediction allow users to search such databases of chemical structures. We present a new strategy called MetFusion to combine identification results from several resources, in particular, from the in silico fragmenter MetFrag with the spectral library MassBank to improve compound identification. We evaluate the performance on a set of 1062 spectra and achieve an improved ranking of the correct compound from rank 28 using MetFrag alone, to rank 7 with MetFusion, even if the correct compound and similar compounds are absent from the spectral library. On the basis of the evaluation, we extrapolate the performance of MetFusion to the KEGG compound database.

Technical Challenges in Mass Spectrometry-Based Metabolomics

“…[71][72][73] Although a rule to account for mass shi s in fragmentations of even-electron organic ions has been proposed, 74) precise prediction of fragments produced by collision induced dissociation of target molecules continues to be challenging in computational mass spectrometry research. 73,[75][76][77][78] Since the databases of naturally occurring metabolites have been developed and enriched by methods such as in silico derivatization of chemical structure, 79,80) the prediction of tandem mass spectra data from the metabolite structure, based on some fragmentation rules, 75) quantum chemistry simulations as well as machine learning techniques 73) would be key technologies for metabolomics and computational mass spectrometry. For example, the competitive fragmentation modeling engine has been developed to produce a probabilistic generative model for the CID fragmentation process by machine learning techniques, which has been helpful in compound identi cation in GC-MS metabolomics.…”

Metabolomic Analysis of Schizosaccharomyces pombe: Sample Preparation, Detection, and Data Interpretation

“…If MS/MS spectra are available, comparison with spectral databases such as MassBank (Horai et al 2010) and METLIN (Smith et al 2005) can also greatly assist with the verification of compound identity. An integrated identification strategy called MetFusion was recently developed (Gerlich and Neumann 2013), combining spectral database searches with in silico fragmentation prediction.…”

Solving CASMI 2013 with MetFrag, MetFusion and MOLGEN-MS/MS

“…9) Additional submissions with di erent strategies were created for this manuscript. To assess the e ect of database joining, candidate lists obtained from individual databases were evaluated separately, while all challenges were re-run with and without element restrictions to assess the formula ltering applied and, especially for Challenges 15 and 16, to generate results retrospectively that contained the correct candidate.…”

“…e size of the challenges (up to 1,440 Da), relatively large ppm errors in some challenges and lack of diagnostic fragment information indicated that structure generation approaches would not be feasible for Category 2. Instead, the challenges favoured the compound database approaches of MetFrag 8) and MetFusion 9) for Category 2, molecular structure identi cation. MetFrag accepts a formula or mass plus accuracy information as an input to query compound databases such as PubChem, 10) ChemSpider 11) and KEGG 12) for candidate structures.…”

“…More details are given in Gerlich and Neumann. 9) For CASMI 2013, the MOLGEN-MS/MS, MetFrag and MetFusion approaches were combined with the aim of markedly improving on the 2012 results 13) and moving towards a work ow-style automated elucidation.…”