dc.description.abstract | Prostate cancer (PCa) remains a predominant health concern worldwide among men, with a critical need for improved biomarkers that can accurately predict disease progression and guide treatment decisions. This thesis provides significant advancements in the understanding of PCa through comprehensive multi-level transcriptomic analyses, utilising microarray, bulk, and single-cell RNA sequencing (scRNA-seq) data, aimed at enhancing diagnostic, risk stratification and prognostic capabilities.
Chapter 3 presents a transcriptome-wide gene expression analysis of histologically confirmed malignant and matched benign prostate samples. This analysis identified a cluster of dysregulated and co-expressed genes associated with PCa. A novel five-gene signature, developed through the integration of public data, demonstrated significant predictive power for progression-free survival (PFS) and effectively stratified patients into distinct risk groups, with high-risk patients exhibiting poorer outcomes. An enhanced nomogram incorporating this signature and the Gleason score outperformed traditional clinicopathological and Cambridge Prognostic Groups (CPG) models in predicting PFS.
Chapter 4 investigates the role of alternative splicing (AS) in PCa, a key factor in transcriptomic diversity and cancer progression. The study identified consistently dysregulated splicing events across multiple cohorts and developed a six-event prognostic signature that effectively distinguished between different risk groups for biochemical recurrence-free survival (BCRFS), with high-risk patients experiencing worse outcomes. This signature, combined with the Gleason score and pathological tumour stage, significantly enhanced BCRFS prognostication, especially for 5-year survival predictions.
Chapter 5 assesses the prognostic utility of epithelial cell marker genes (ECMGs) in PCa by integrating bulk and scRNA-seq data. The study yielded a robust 11-ECMG-based risk signature that effectively categorised patients into distinct BCRFS risk groups, with high-risk patients showing poorer outcomes. This signature showed superior overall predictive accuracy compared to published PCa signatures and demonstrated its value as an independent prognostic factor, particularly in larger cohorts. Additionally, the exploration of blood-based datasets identified several ECMG signature genes that hold promise as non-invasive biomarkers, potentially improving the monitoring of treatment efficacy and overall survival in PCa patients.
Overall, this thesis underscores the complexity of PCa pathology and the potential of advanced transcriptomic profiling paired with machine learning to improve prognosis accuracy. The gene signatures and nomograms developed here may offer significant potential for refining prognosis, enhancing clinical management and improving patient outcomes in PCa. | en_GB |