A Clinically Adjusted Transcriptomic Signature for Survival Stratification in High-Grade Serous Ovarian Cancer
Aicha Lakhssassi, Mohcine Bennani Mechita, Joaira BakkachPURPOSE
High-grade serous ovarian cancer (HGSOC) remains the most lethal gynecologic malignancy, and clinical outcomes vary substantially even among patients with apparently favorable surgical cytoreduction. Improved molecular risk stratification may help identify patients at higher risk despite standard clinical prognostic features. Using publicly available transcriptomic and clinical data, we developed a compact transcriptomic signature for overall survival and assessed whether it added prognostic information beyond age, stage, and residual disease.
METHODS
Using TCGA-OV RNA-seq and clinical data (n = 416 HGSOC), we built a parsimonious survival signature by variance filtering, univariable Cox screening, 5-fold cross-validated LASSO-Cox, and multivariable Cox pruning. Performance was assessed on a held-out 30% test partition, in multivariable Cox adjusted for age, stage, and residual disease, and within debulking strata. Robustness was tested by 100-bootstrap resampling and consensus across Random Forest, Gradient Boosting, and L1-logistic regression. Differential expression in two additional HGSOC cohorts (n = 165 tumors) supported biological interpretation.
RESULTS
A nine-gene signature emerged: PRSS16, CTXN2, FCRLA (protective); PI3, LILRA2, LCN6, RNF103-CHMP3, KLK3, MTRNR2L8 (adverse). Held-out test: HR 2.15 (95% CI 1.35–3.43, p = 1.2×10
-3
); 5-year time-dependent AUC 0.69. In multivariable Cox (n = 363), the signature was independently associated with survival (HR 1.76 per SD,
CONCLUSION
This exploratory nine-gene transcriptomic signature, derived from publicly available HGSOC transcriptomic and clinical data, stratifies overall survival and adds prognostic information beyond stage and surgical debulking. Its retained value among optimally debulked patients addresses a clinically important setting where additional risk stratification is needed. The signature reflects established TLS–mesenchymal tumor biology and nominates relatively understudied components, including CTXN2, RNF103-CHMP3, and MTRNR2L8, for orthogonal validation. External survival validation is required before clinical translation.