Computational Methods for Canonical and Noncanonical Peptide Identification in Immunopeptidomics
by John Alexander Cormican
Date of Examination:2025-04-11
Date of issue:2025-05-19
Advisor:Dr. Juliane Liepe
Referee:Prof. Dr. Matthias Dobbelstein
Referee:Dr. Johannes Soeding
Referee:Prof. Dr. Anne-Christin Hauschild
Referee:Prof. Dr. Lutz Walter
Referee:Prof. Dr. Jürgen Wienands
Persistent Address:
http://dx.doi.org/10.53846/goediss-11246
Files in this item
Name:thesis_john_cormican_final.pdf
Size:17.4Mb
Format:PDF
This file will be freely accessible after 2026-04-10.
Abstract
English
The presentation of peptides via the MHC-I pathway as signals to CD8+ T cells is an essential element of the adaptive immune system. Hence, effective identification of the peptides presented (i.e., the MHC-I immunopeptidome) advances basic science while having immediate translational application in medicine. From a basic science perspective, reliable identification of the MHC-I immunopeptidome sheds light on a fundamental pathway within the cell. The translational applications have been particularly highlighted by the recent Covid-19 pandemic, and the corresponding advances in mRNA-based vaccine delivery provide evidence for continued importance. Immunopeptidome identification is typically achieved via mass spectrometry (MS). However, identification is significantly more challenging than in standard proteomic experiments, necessitating optimized computational solutions. This is the case even in the identification of canonical peptides (peptides that can be traced to the annotated proteome via standard cleavage of a given protein into peptide fragments). These challenges are further exacerbated when we seek to identify noncanonical peptides – sequences which are not found within the annotated proteome and may be produced by non-standard transcriptional, translational, or post-translational processing. These processes include translation of sections of the genome thought to be non-coding as well the phenomenon of proteasome catalyzed peptide splicing (PCPS). The work presented in this thesis focusses on peptide identification with special focus on optimizing noncanonical peptide identification. We began by laying the foundations, by developing two software tools which were at the heart of all further analysis. Firstly, the iBench software allowed rigorous benchmarking of peptide identification tools, hence validating all further advances. The other foundational software is the inSPIRE tool. inSPIRE combines multiple advances in machine learning to provide highly sensitive identification of canonical peptides in the MHC-I immunopeptidome. From this point, we moved quickly to application, firstly with noncanonical peptide identification in an in vitro setting. In our first application focused study, a modified version of inSPIRE was leveraged to identify peptides produced via PCPS (spliced peptides). This study was primarily focused on furthering our understanding of proteasomal function and the amino acid level preferences driving both cleavage and splicing events. In a second application, the PEPSeek tool was developed to extract pathogen derived peptides from MHC-I immunopeptidomics experiments of pathogen infected cells. Again, PEPSeek leveraged inSPIRE for MS identification while also enabling quantitative analysis of immunopeptidomic changes brought about by pathogen infection. This study demonstrated translational value as peptides identifiable only via our software were found to trigger CD8+ T cell response for peptides from both SARS-CoV-2 and listeria monocytogenes. In this study, we also developed the interact-ms web server, further developed in later projects, which allowed the tools developed to be accessed by the wider scientific community. Finally, in the cornerstone work of this thesis, we developed a new identification software (PISCES) for the identification of noncanonical peptides in the MHC-I immunopeptidome. This tool allowed reliable and sensitive identification of noncanonical peptides. Large scale application of this software enabled the first integrated analysis of the spliced and cryptic immunopeptidome. Further analysis revealed factors driving both spliced and cryptic peptide presentation. Strong correlation to our in vitro work speaks to the validity of our results in both studies. Thus, through the development of novel computational methods, we provide a reliable and more complete picture of the canonical and noncanonical immunopeptidome. Further analysis of these results provided biological and biochemical insights into the processes generating these peptides. In conclusion, the work presented in this thesis (i) improves the sensitivity of standard canonical peptide identification, particularly in the MHC-I immunopeptidome, (ii) shows the translational relevance of the improvements presented, (iii) ensures that the advances are available to the wider scientific community, (iv) enable noncanonical peptide identification in the immunopeptidome, and (v) sheds light on the mechanisms behind noncanonical peptide generation.
Keywords: mass spectrometry; pcps; immunopeptidomics; epitope discovery; computational tools; peptide identification
