Quantum Chemical Prediction of Electron Ionization Mass Spectra of Trimethylsilylated Metabolites

Chemical
derivatization, especially silylation, is widely used
in gas chromatography coupled to mass spectrometry (GC-MS). By introducing
the trimethylsilyl (TMS) group to substitute active hydrogens in the
molecule, thermostable volatile compounds are created that can be
easily analyzed. While large GC-MS libraries are available, the number
of spectra for TMS-derivatized compounds is comparatively small. In
addition, many metabolites cannot be purchased to produce authentic
library spectra. Therefore, computationally generated in silico mass
spectral databases need to take TMS derivatizations into account for
metabolomics. The quantum chemistry method QCEIMS is an automatic
method to generate electron ionization (EI) mass spectra directly
from compound structures. To evaluate the performance of the QCEIMS
method for TMS-derivatized compounds, we chose 816 trimethylsilyl
derivatives of organic acids, alcohols, amides, amines, and thiols
to compare in silico-generated spectra against the experimental EI
mass spectra from the NIST17 library. Overall, in silico spectra showed
a weighted dot score similarity (1000 is maximum) of 635 compared
to the NIST17 experimental spectra. Aromatic compounds yielded a better
prediction accuracy with an average similarity score of 808, while
oxygen-containing molecules showed lower accuracy with only an average
score of 609. Such similarity scores are useful for annotation of
small molecules in untargeted GC-MS-based metabolomics, suggesting
that QCEIMS methods can be extended to compounds that are not present
in experimental databases. Despite this overall success, 37% of all
experimentally observed ions were not found in QCEIMS predictions.
We investigated QCEIMS trajectories in detail and found missed fragmentations
in specific rearrangement reactions. Such findings open the way forward
for future improvements to the QCEIMS software.