Author: Rastko Stojšin, MS (2026)
Primary advisor: Hongfang Liu, PhD
Committee members: Arif Harmanci, PhD, Degui Zhi, PhD and Jinlian Wang, PhD
PhD thesis, McWilliams School of Biomedical Informatics at UTHealth Houston.
ABSTRACT
Post-translational modifications (PTMs) are chemical modifications that occur on proteins after synthesis and play an essential role in regulating protein function. Dysregulation of biologically critical PTMs has been widely implicated in many disease phenotypes primarily through effects on complex downstream protein networks. Measurement of PTMs—particularly abundant and rapidly cycling modifications—is often limited to technically challenging and low-throughput biochemical assays. As a result, large-scale clinical datasets rarely contain direct measurements of PTM activity. These limitations restrict current research to expensive, small-scale, hypothesis-driven laboratory studies often focused on an individual disease and pathway.
This dissertation presents a data-driven biomedical informatics framework for quantifying dysregulation of tightly regulated PTM systems and identifying downstream molecular pathways through which this dysregulation contributes to disease.
The framework integrates publicly available large-scale transcriptomic, clinical, and molecular data to overcome key experimental and data limitations allowing for scalable research in PTM dysregulation and systems-level understanding of disease progression. Disease-associated dysregulation of tightly regulated PTM systems is inferred by modeling coordinated expression patterns among their regulatory components. Functional networks of downstream genes are built, and network-aware computational approaches identify pathways that differentiate clinically relevant cohorts.
The approach is broadly applicable to tightly regulated disease-associated PTMs, but the framework was developed and evaluated using the cancer-associated PTM, O-GlcNAcylation as the model regulatory system. Reliable and biologically relevant quantification of O-GlcNAcylation dysregulation was demonstrated across six different cancer types and validated using independent external dataset. The network-based analysis identified biologically meaningful downstream pathways evaluated through cross-cohort comparisons, pathway enrichment analyses, and associations with clinical outcomes. The identified subnetworks show strong enrichment for known oncogenic pathways and exhibit significant associations with clinical outcomes across several cancer types.
These results establish a generalizable biomedical informatics framework for studying PTM dysregulation and downstream phenotypic effects without requiring direct experimental measurement.