Significance Analysis of Prognostic Signatures Applied to Prostate Cancer
Laleh Montaser, Andrew H Beck. Beth Israel Deaconess Medical Center, Boston, MA
Background: The TMPRSS2: ERG gene fusion is present in ∼50% of prostate cancers. Little is known of biological pathways driving disease progression in prostate cancer stratified by ERG fusion status. We recently developed Significance Analysis of Prognostic Signatures (SAPS), a computational method for identifying statistically significant prognostic biological signatures in clinically annotated genomic data sets. To identify biological pathways driving disease progression in prostate cancer, we used SAPS to perform a meta-analysis of prognostic pathways in ERG-positive and ERG-negative prostate cancer.
Design: We identified 2 large data-sets (GSE10645 and GSE8402)with clinically annotated expression profiling data in the Gene Expression Omnibus. Using the GSE8402 data-set (which contains FISH-based data on ERG fusion status, we built an expression-based model to predict fusion status. The model was applied to all cases to classify them as ERG+ or ERG-. We performed a subtype-specific data-normalization to transform expression values into a Z score within ERG+ or ERG- cases. The normalized values were merged and SAPS was applied separately to ERG+ and ERG- prostate cancer. The SAPS method computes 3 significance tests for each gene set to indicate the statistical significance of the gene set's prognostic association. The 3 p values are summarized by taking the maximum. We used SAPS to evaluate the prognostic significance of 3,887 biological signatures in ERG+ and ERG- prostate cancer.
Results: This analysis identifies prognostic pathways in both ERG-Positive and ERG-Negative prostate cancer. Using a SAPS score cut-off of 0.10, we identify 34 significant prognostic pathways in ERG+ alone, 4 significant prognostic pathways in ERG- alone, and 1 prognostic pathway significant in both ERG+ and ERG- negative prostate cancer. In ERG positive tumors, altered expression of genes relating to DNA repair, proliferation, and progesterone signaling were significantly correlated with prognosis. In ERG negative cases, gene sets associated with the CHEK2 signalling network and targets of MYC were associated with prognosis. A metastasis-associated gene set was associated with prognosis in both groups.
Conclusions: SAPS is a powerful method for identifying prognostic biological signatures in clinically annotated genomic data sets. Using SAPS on a large prostate cancer meta-data set stratified by ERG status, we identified biological signatures associated with survival in prostate cancer molecular subtypes. These pathways reprsent candidate prognostic signatures and drug targets for ERG+ and/or ERG- prostate cancer.
Category: Genitourinary (including renal tumors)
Monday, March 4, 2013 11:30 AM
Proffered Papers: Section A, Monday Morning