Integrative Analysis of Gene and Protein Expression Data Reveals Novel Clusters for Ovarian Cancer Prognosis
TL;DR
Integrative analysis of multi–omics data provided robust prognostic stratification in ovarian cancer while capturing the underlying molecular and biological features.
Abstract
Ovarian cancer exhibits clinical heterogeneity; reliable prognostic stratification remains challenging despite extensive biomarker research. We performed an integrative analysis of transcriptomic and proteomic data to identify prognostically distinct molecular clusters in ovarian cancer. Using The Cancer Genome Atlas dataset, RNA sequencing and reverse–phase protein array data from 146 overlapping samples were analyzed. Genes (n = 150) and proteins (n = 67) showing nominal associations with overall survival (p < 0.05) were integrated; k–means clustering identified two patient clusters with highly significant differences in survival (p = 8.59 × 10−15). The SAR2 and ACTN4 genes and their corresponding protein expression levels showed high correlations among the significant genes and proteins (correlation coefficients > 0.80). Gene set enrichment analysis revealed that genes regulated by the PAX2 and BCL6 transcription factors were significantly enriched. The identified clusters were validated across six independent public transcriptomic datasets comprising 617 patients; consistent survival separation was observed between the patient clusters (p < 0.05). Clustering based on immunohistochemistry–derived expression profiles of nine proteins in an independent patient cohort demonstrated a significant difference in survival (p = 0.04). Integrative analysis of multi–omics data provided robust prognostic stratification in ovarian cancer while capturing the underlying molecular and biological features.