Identification of a gene signature that can predict lone-term survival in patients with high-grade late stage serous ovarian cancer

Abstract

5032 Background: Though most patients with late stage serous ovarian cancer die within 2 years of diagnosis, a subset of patients, even with clinically and morphologically indistinguishable diseases, develop a more chronic form of ovarian cancer, and may survive 5 years or more with treatment. We hypothesize that these cancers may have a prognostic signature that can be used to predict the clinical course of the disease. Methods: To discover the prognostic signature, expression profiling was performed on 53 microdissected high-grade late stage serous tumors using the Affymetrix U133 plus2 oligo array platform. We applied a “semi-supervised” dimension reduction method, and (1) identified the genes associated with cancer survival and (2) established a predictive model. In brief, univariate Cox regression with Jackknife safeguard were conducted and pinpointed the 200 most significant genes. The first 6 principal components of these genes, representing >90% of total variance, entered a multivariate Cox model, based on which the relative hazard of future patients can be predicted. To confirm our finding, the microarray data underwent leave-one-out validation. In each iteration, we reserved one sample, and used the remaining 52 patients to carry out the gene selection and model building, which furthermore predicted the relative hazard of the reserved patient. To further validate the gene signature, quantitative PCR was performed on the 10 most significant genes in 44 patients. Results: According to the prediction, the patients were equally divided into low- and high- risk groups and the non-parametric Kaplan-Meier plot and log rank test showed the two groups were significantly different in survival (p = 0.0089). Quantitative PCR validation demonstrated the accuracy and robustness of the prognostic gene signature and the prediction model. Conclusions: These genes contribute to the differential survival of ovarian cancer patients, and may serve as unique targets for the management of the disease. No significant financial relationships to disclose.