Identification of Prostate Cancer Recurrence Signatures Using Unsupervised Machine Learning
Authors
Rifai, Safiullah ; Rifai, Azimullah ; Meher, Zumar ; Khan, Mohammad ; Wang, Linbo ; Guang, Wei ; Hussain, Arif, M.D.
Advisor
Date
Embargo until
Language
Book title
Publisher
Peer Reviewed
Type
Research Area
Jurisdiction
Collections
Other Titles
See at
Abstract
Radical prostatectomy or radiation are the standard treatment for clinically localized prostate cancer. Although potentially curative, up to 15-30% of men undergoing definitive local therapy experience biochemical recurrence. The time to biochemical recurrence can serve as an important indicator of worse long term clinical outcomes. Therefore, understanding the factors driving early recurrence is critical. Our study examined transcriptomic data from 74 patients in the TCGA Prostate Adenocarcinoma dataset who experienced biochemical recurrence after radical prostatectomy for localized disease. We stratified patients into quartiles based on time to recurrence and analyzed their gene expression profiles using K-means clustering and Principal Component Analysis (PCA). K-means clustering revealed distinct gene expression patterns associated with earlier biochemical recurrence. Over-representation analysis across multiple Gene Ontology collections highlighted immune cell signatures as major contributors to these patterns. Principal Component Analysis identified epithelial splicing regulatory protein 1 (ESRP1) – a regulator of epithelial-mesenchymal transition – as significantly associated with early recurrence in a subset of patients. Future directions include deconvolution of the bulk RNA sequencing data to better characterize immune signatures across recurrence quartiles, and extension of our analysis to include methylation and protein expression data.
