Research And Application Of Protein Theoretical Spectrum And Peptide Fragmentation Event Model

The protein identification based on tandem mass spectrum depends on an accurate prediction of theoretical spectrum,which in turn heavily depends on a quantitative understanding of the peptide fragmentation process in mass spectrometry.Due to the lack of the knowledge on peptide fragmentation,the widely used database searching methods up to date,such as SEQUEST and Mascot,adopted a simple model without considering some important factors influencing fragmentation,such as the position and the type of the peptide bond to be broken.It has been reported that the ignorance of above important factors usually leads to a theoretical spectrum with significant deviation.In addition, experimental spectrum usually contain many ions with neutral losses,such as dehydration and deamination.How to quantify the intensity of these ions will greatly affect the prediction of theoretical spectrum.In this thesis,we present the following attempts to tackle the above problems:(1) We proposed a new peptide fragmentation model.In the field of mass spectrum,whether a b ion or ay ion will be generated in a fragmentation remains a puzzle.A widely used trick is to set b/y ratio to be 1:1,which is obviously biased from the observation.To overcome this difficulty,we proposed "fragmentation event model";that is,we predict the possibility of fragmentations at different peptide bonds rather than intensity of ions.In particular,we quantify the influence of both position and type of the peptide bonds based on a training set of MS/MS spectra.(2) We proposed a measurement to evaluate the similarity between the experiment spectrum and the candidate peptide sequence.We first convert experimental spectrum into a fragmentation event distribution,and then predict theoretical fragmentation event distribution for a given sequence.Jensen-Shannon Divergence are employed to measure the similarity between these two event distributions.(3) We designed an algorithm to derive the neutral loss possibilities of ions in the peptide fragmentation process.Neutral loss is an important fragmentation pathway in peptide gas phase fragmentation reaction,and understanding this phenomenon contributes significantly to the peptide fragmentation mechanism and protein identification through database search.Given the current insufficient understanding of neutral loss mechanism and its important role in peptide sequencing, we made efforts to derive neutral loss probabilities and incorporate the scheme into peptide sequencing algorithm to predict more realistic spectrum for peptide.We implemented the above algorithms into an open source package PI(Peptide Identification).Experimental results on different data sets from several kinds of spectrometers, e.g.,QTOF,Ion-Trap,LTQ-FT,etc.,suggest that our fragmentation event model along with Jensen-Shannon Divergence has advantages over SEQUEST and MASCOT.The neutral losses probabilities can be used to improve theoretical spectrum prediction.Furthermore,experimental results also provide quantitative supports of the observations regarding fragmentation and neutral losses. iteration algorithm,EM algorithm...