Incorporating genotype information in a precise prediction model for platinum sensitivity in epithelial ovarian cancer.

Objective: Develop a predicting model that can help stratify patients with epithelial ovarian cancer (EOC) before platinum-based chemotherapy. Methods: 148 patients with pathologically confirmed EOC and with a minimum 5-year follow-up were retrospectively enrolled. Patients were classified into platinum-sensitive and platinum-resistant groups according to treatment responses. The correlation between clinical factors and drug sensitivity was evaluated using statistical tests. Approximately 700,000 single-nucleotide polymorphism (SNP) sites were assessed for association with drug sensitivity via the Genome-wide Association Study (GWAS). LASSO regression and manual selection were employed to reduce the number of variables. A predicting model based on optimized variables was constructed. The predictive ability of the model was assessed using the Kaplan-Meier curve. Results: No statistically significant association was found between clinical factors and drug sensitivity. Sixteen SNPs were preserved after the optimization. A predicting model for drug sensitivity was constructed based on those sixteen SNPs. Coefficients of the synergistic effect for each SNP were determined, and an algorithm of the Drug Sensitivity Index (DSI) was built. The DSI score can successfully distinguish the drug-sensitive or drug-resistant patients with sensitivity, specificity, positive predictive value, and accuracy of 94.7%, 83.3%, 90.8%, and 90.5%, respectively. In both the training set and validating samples, the Kaplan-Meier curve showed that the median PFS and mean OS were significantly differentiated between the predicted sensitive and resistant patients ( Conclusions: A mathematical model incorporating genotype information could help predict the drug sensitivity of platinum-based chemotherapy before the treatment in EOC patients. A personal chemotherapy could be achieved based on the model.