Summary: | <p>Tandem mass spectrometry (MSMS) has become a powerful tool for the analysis
of biomolecules. To reveal molecular identity, experimental mass (<I>m/z</I>) data are either
matched to appropriate databases or processed <I>de-novo</I>. Both approaches are essentially
one-dimensional, because the <I>m/z</I> values of fragment ions play dominant role, while the
intensities (the second dimension) are being neglected, despite their potential to
corroborate or contradict the identification. Unlike the trivial case of <I>m/z</I> values,
predicting the ion intensities has not been mastered yet beyond empirical observation and
statistical treatment. This dissertation presents a fundamental, structure-based algorithm
for the prediction of fragment ion intensities in MSMS spectra of peptides and
metabolites.</p>
<p>The algorithm builds on the central hypothesis that the fragment ion intensities
reflect the relative abundances of respective protonated precursor isomers prior to
fragmentation. The hypothesis is supported by extensive experimental evidence showing
that in ion trap mass detectors, relative ion intensities do not depend on the energy or
activation time of fragmentation when commonly used ranges of conditions are explored.</p>
<p>The multi-step algorithm developed includes molecular mechanics Monte Carlo
conformational space sampling, semi-empirical calculations, and Density Functional
Theory (DFT) quantum chemistry computations for structure refinement and energy
calculations. A Boltzmann distribution determined from energy values of pertinent
precursors accurately corresponded with relative ion intensities in MSMS spectra of three
model pentapeptides.</p>
<p>Reproducibility of the algorithm was tested, and while substantial differences
were revealed among the multiple Monte Carlo samplings started from the same initial
structures, the inconsistency was mitigated in following semi-empirical and DFT steps.</p>
<p>The algorithm was optimized for efficiency as well. Computational costs were
lowered (by more than 50%) by narrowing the energy window in which the conformers
were taken to the following steps in the algorithm and finally to the Boltzmann
distribution.</p>
<p>For metabolites, ion intensity orders for nine out of eleven molecules were
predicted correctly. However, the accuracy of the prediction of relative ion intensities
was not satisfactory. Nevertheless, predicting the most intensive ion alone could be
invaluable for preliminary metabolite identification and selecting good candidate
standards for ultimate identification based on matched properties of analyte and standard.</p>
|