DATA MINING OF PEPTIDE MS/MS SPECTRA TO ELUCIDATE GAS-PHASE PEPTIDE DISSOCIATION MECHANISMS AND IMPROVE PROTEIN IDENTIFICATION

Abstract

Mining of datasets obtained from proteomics experiments was performed to investigate the dissociation pathways of protonated peptides activated in the gas phase under low energy collision-induced dissociation (CID). Intensity patterns in ion trap tandem mass spectra were exploited and different statistical approaches were employed to elucidate the dissociation mechanisms.Chapter 2 describes a study of 506 doubly-protonated tryptic peptides that shows the presence of an internal basic residue can increase the preferential fragmentation C-terminal to aspartic acid (Asp-Xxx) significantly. The degree of enhancement varies with the identity of the basic residues. The result corroborates a previously published mechanism based on studies from model peptides, and was incorporated into an existing peptide sequencing algorithm. A preliminary test on a separate dataset of 119 spectra shows that implementing rules to predict enhanced cleavages at Asp-Xxx improves the ability of the algorithm to identify the correct sequence from a list of candidates.Chapters 3-4 describe much more elaborate analyses on 28,330 peptides of different sequences and charge states. Extensive sorting based on prior knowledge was first performed to probe the correlation of fragmentation patterns with structural features. Pair-wise fragmentation maps reveal that the difference in basicity between Arg and Lys results in different dissociation patterns among singly-protonated tryptic peptides. While one dominant protonation form (proton localized) exists for Arg-ending peptides, a heterogeneous population of two or more protonated forms (proton partially-mobile) exist for Lys-ending peptides. Asp/Glu-Xxx dominates spectra from peptides that have a localized proton(s) and Xxx-Pro dominates those that have a mobile or partially mobile proton(s). When Pro is absent from peptides that have a mobile or partially mobile proton(s), cleavage at each peptide bond becomes more prominent. A fundamental dependence of gas phase peptide fragmentation on conformational constraints was found.A knowledge mining scheme was proposed in Chapter 5 to bypass the prior knowledge constraints and cluster the dissociation behaviors of 28,330 peptides into four distinct categories. The most influential factors in the fragmentation process are: the mobility of the proton(s), the presence and the location of Pro and Arg. Structural motifs responsible for each dissociation behavior are also elucidated.