A database searching approach can be used for metabo
lite identification in metabolomics by matching measured tandem mass spectra (MS/MS) against the predicted fragments of metabo
lites in a database. Here, we present the open-source MIDAS algorithm (Metabo
lite Identification via Database Searching). To evaluate a metabo
lite-spectrum match (MSM), MIDAS first enumerates possible fragments from a metabo
lite by systematic bond dissociation, then calculates the plausibi
lity of the fragments based on their fragmentation pathways, and finally scores the MSM to assess how well the experimental MS/MS spectrum from col
lision-induced dissociation (CID) is explained by the metabo
lite鈥檚 predicted CID MS/MS spectrum. MIDAS was designed to search high-resolution tandem mass spectra acquired on time-of-f
light or Orbitrap mass spectrometer against a metabo
lite database in an automated and high-throughput manner. The accuracy of metabo
lite identification by MIDAS was benchmarked using four sets of standard tandem mass spectra from MassBank. On average, for 77% of original spectra and 84% of composite spectra, MIDAS correctly ranked the true compounds as the first MSMs out of all MetaCyc metabo
lites as decoys. MIDAS correctly identified 46% more original spectra and 59% more composite spectra at the first MSMs than an existing database-searching algorithm, MetFrag. MIDAS was showcased by searching a pub
lished real-world measurement of a metabolome from
Synechococcus sp. PCC 7002 against the MetaCyc metabo
lite database. MIDAS identified many metabo
lites missed in the previous study. MIDAS identifications should be considered only as candidate metabo
lites, which need to be confirmed using standard compounds. To faci
litate manual va
lidation, MIDAS provides annotated spectra for MSMs and labels observed mass spectral peaks with predicted fragments. The database searching and manual va
lidation can be performed on
line at
http://midas.omicsbio.org.