Computational Methods for Characterizing Post-translational and Chemical Modifications Found in Open Searches

Abstract

Post-translational modifications (PTMs) govern many processes within cells and understanding their function is critical to both the basic and biomedical sciences. However, identifying modified peptides, particularly unexpected and rare modifications, remains a challenge to proteomics researchers. Recent advances in proteomics search tools have expanded the capacity to identify the entire modification landscape in an unbiased way, but the modifications identified in this manner—called “open searching”—require extensive post-processing to elucidate their identities. In this dissertation, I develop computational methods to characterize and identify modifications derived from open searches.

In Chapter Two, I develop a method for comprehensively characterizing open search results, PTM-Shepherd, enabling new applications for unbiased PTM discovery. PTM-Shepherd automates characterization of PTM profiles detected in open searches based on attributes such as amino acid localization, fragmentation spectra similarity, retention time shifts, and relative modification rates. I show how open searches can be used to profile experimental artifacts by identifying a set of PTMs common across several formalin-fixed paraffin-embedded datasets that researchers can include in future analyses, identifying a range of Cys-specific artifacts in a commonly used high-quality dataset, finding two previously undescribed PTMs in synthetic peptide data and TMT data, and tracing major site-specific PTM batch effects in a multi-university consortium’s proteomics data back to sample processing. In Chapter Three, I extend the algorithm developed in Chapter Two, introducing additional metrics that allow researchers to peer into the spectra of PTMs and extract PTM-specific fragmentation patterns and diagnostic ions. I find new diagnostic for multiple common PTMs, including identifying new fragmentation patterns for glycopeptides under high energy fragmentation, new diagnostic ions for sialic acid under high energy fragmentation, and new diagnostic ions and peptide remainder masses for ADP-ribosylation, as well as examining general trends in the utility of PTM-specific diagnostic features such as the inverse relationship between an ion’s average intensity and its specificity to the modification. In Chapter Four, I expound my methods’ utility by applying it in multiple settings to characterize synthetic and chemical PTMs. In each case, I show how these methods aid in interpretation of results or increase coverage of the proteome by recovering additional modified peptides. For chemoproteomics probes, I demonstrate how expensive isotopic labeling to identify fragmentation patterns can be avoided, finding multiple novel diagnostic ions for a Cys-specific triazole biotin probe. For RNA crosslinked data, I show how the number of recovered identifications increases by up to 50% over existing state of the art methods when incorporating fragmentation information discovered by PTM-Shepherd. Finally, for protein-protein crosslinking, I show how PTM-Shepherd can derive fragmentation patterns for non-cleavable crosslinkers without computationally expensive or custom workflows, discovering that auto crosslinks can be used to identify fragment remainder masses for that can reduce computational complexity during searching.

The ability to survey the entire post-translational modification landscape has major implications across proteomics subdisciplines. In total, the work described herein represents a major milestone in the interpretation of open search results and opens the door to better understandings of cellular processes and disease by facilitating new modes of analysis.