Objective: Prognoses of ovarian cancer (OC) have improved with the paclitaxel-carboplatin regimen. However, it remains unclear which cases exhibit a genuine benefit from taxane or from platinum. We aimed to predict taxane and platinum sensitivity in OC via gene expression.
Methods: We identified differentially expressed genes in responsive and resistant cases from advanced OC biopsy expression dataset GSE15622, containing responses to paclitaxel or carboplatin monotherapy. These genes generated a scoring system for prediction of drug response by applying single-sample gene set enrichment analysis. Discriminative metrics termed the T-score and C-score were derived.
Results: High C-score levels were significant in responders compared to non-responders in a separate cisplatin treatment dataset (GSE18864, p=0.043). High C-score groups also had significantly better progression-free survival in three OC datasets (The Cancer Genome Atlas, TCGA: p=0.02; GSE9891: p=0.03; GSE30161: p=0.001). In two additional datasets of advanced OC, high T-scores could associate taxane and platinum regimens with better survival than non-taxane and platinum regimens (GSE9891: p<0.0001; GSE3149: p=0.045), whereas in cases with low T-scores, different chemotherapeutic regimens did not result in a significant difference. Assessing TCGA and GSE9891, T-scores were elevated in the C1/Mesenchymal subtype, whereas C-scores were elevated in the C5/Proliferative subtype and were lower in the C1/Mesenchymal subtype (p<0.0001, respectively). C- and T-scores negatively correlated with each other, suggesting complementary roles of taxane and platinum.
Conclusions: Our proposal and finding of a scoring system that could predict platinum or taxane response could be useful to develop individualized treatments to ovarian cancer.
Keywords: Gene expression profile; Ovarian cancer; Platinum; Prediction; Sensitivity; Taxane.
Copyright © 2016 Elsevier Inc. All rights reserved.