Background: Identification of detected compounds in untargeted
LC/MS profiling is a common bottleneck in metabolomics. The CASMI contest
challenges mass spectrometry experts and algorithm developers to evaluate how
reliable their methods derive molecular formulae and structures from blinded mass
Objective: The application of the MAGMa software to solve the CASMI 2014 challenges
Methods: MAGMa was used to automatically retrieve candidate molecular structures
from the HMDB and PubChem chemical databases, based on MS1 precursor
m/z values, and to provide a score indicating how well they explain the accurate
Results: For 40 out of 48 challenges, candidates with the correct molecular formula were ranked on
top. For 22 out of 42 challenges the top-ranked candidate also represented the correct chemical structure
and in 9 other cases the correct molecule was ranked in the top 10.
Conclusion: Advantages and limitations of the approach and consequences with respect to retrieving
and scoring of the correct candidates are discussed.