Вычислительные методы для аннотирования данных тандемной масс-спектрометрииComputational methods for tandem mass spectrometry data annotation

Tandem mass spectrometry has become the standard method to identify molecules and proteins in complex biological mixtures (e.g. blood drops), and the data obtained is routinely analyzed with computer programs. Roughly, 50-80% of the data cannot be annotated with high confidence due the following issues: (1) low discriminative power of the score functions, i.e. the ability to discriminate between correct and incorrect annotations, (2) raw scores are not calibrated properly and lack universal semantics, (3) incomplete candidate peptide sets, i.e. the correct peptide is not included to the reference set possibly due to a post-translational modification (PTMs), (4) noise, i.e. the mass spectrometer produces considerably amount of noise in the spectra, and (5) detector/Instrument accuracy. The dissertation presents novel computational methods to tackle down the issues 1-4 and they achieved state-of-the-art performance in data annotation at the time of their publication. Therefore, more accurate data identification and spectrum annotation can be obtained which and experimenters and practitioners can conclude correct conclusions about their experiments and to make proper decisions for future experiments or clinical therapies, for instance, in selecting the right drug therapy.