Fig. 6: Transcriptomic characteristics associated with lack of clinical benefit and development of an NCB signature.

a Gene set enrichment analysis of pretreatment RNAseq identifies pathways enriched in NCB versus CB tumors across both treatment arms, including vascular adaptation, developmental morphogenesis, and cytoskeletal remodeling. Pathways upregulated in NCB with |NES| >1 and padj <0.05 (Benjamini–Hochberg method) are shown. Circle size reflects leading-edge gene count. b NCB signature scores (18-gene overlap derived from shared cytoskeletal and morphogenesis pathways) in CB versus NCB tumors in D + O + C, D + C, and GSE206422 datasets (recurrent ovarian cancers treated with pembrolizumab, bevacizumab, and metronomic cyclophosphamide). Scores were calculated by singscore, and Wilcoxon rank-sum tests (two-sided) were used for comparisons. Box plots display the full range (min to max), with the median indicated by a horizontal line. c Spearman correlation between NCB scores and PFS. The solid line represents the best-fit linear regression, and the shaded area indicates the 95% CI of the regression. d Kaplan–Meier estimates comparing high versus low NCB score groups, with the hazard ratio, with 95% CI and log-rank P value. Pretreatment RNAseq data for a–d were available for D + O + C (21 CB including four exceptional responders; 10 NCB; n = 31), D + C (14 CB including three exceptional responders; 10 NCB; n = 24), and GSE206422 (21 CB including six exceptional responders; 6 NCB; n = 27). Source data are provided as a Source Data file. CB clinical benefit, CI confidence interval, D + O + C durvalumab, cediranib, and olaparib, D + C durvalumab plus cediranib, NCB no clinical benefit, NES normalized enrichment score, padj adjusted p value, PFS progression-free survival.