Frequently deleted genes in ovarian cancer as indicators of platinum response.

Abstract

5583 Background: Integrating chromosomal deletions/amplifications with sequencing alterations is increasingly important in the determination of key drivers of outcome in cancer (Leary 2008, Curtis 2012). We introduce a novel computational approach, Gene Set Outcome Analysis, to determine gene signatures constrained to regions with frequent deletion or amplification events in ovarian cancer identified by TCGA. Differential expressions of mRNA from these genes are used as a proxy for loss of function from deletions or amplifications. Methods: Gene signatures from each region were evaluated using log-rank test comparing high and low gene expression groups split by cohort mean. In total, 30,119,708 signatures constrained to 47 deleted and 36 amplified regions were tested for PFS for first line platinum treatment in a random 2/3 split of TCGA ovarian cohort (n=262) resulting in 26,271 gene signatures with p < .001. The remaining 1/3 cohort (n=135) and full cohort (n=397) were used as a replication study where 111 signatures have a p < .01 located in 2 deletion and 1 amplification region. The signature with the lowest p-value from each region that validated in the replication study, were evaluated in GSE9891 (n=179) (Tothill 2008) for PFS when treated with platinum and taxol resulting in gene signatures from 2 deletion regions with p<.05. Results: Greater than mean expression in this gene signature [OXTR, SATB1, WNT7A, SH3BP5, CRBN, ATG7, CRELD1, TMEM43] located at 3p23-p26.2 predicts poor response to platinum chemotherapy in TCGA full ovarian cohort (17.9 vs 14 months p = 1.4E-7) and validates in GSE9891 (19.6 vs 13.3 months p = .014). Using a separate, compact gene signature [CAAP1, LRRC19, IFNA8] located at 9p21.2, samples with lower than mean expression demonstrate increased platinum sensitivity in TCGA full ovarian cohort (12.2 vs 18.2 months p = 1.0E-6) and in GSE9891 (15.5 vs 18.6 p = .055). Conclusions: Response to platinum therapy is an important predictor of survival in high-risk ovarian cancer patients. These signatures arising from common deletion and amplification events in ovarian cancer can anticipate platinum sensitivity and merits further study for use in choosing optimal therapies studying platinum resistance mechanisms.