Multi-Cohort Transcriptomic Analysis Reveals Heterogeneous Ferroptosis-Related Transcriptional Programs Across EAOC-Associated Pathological States

Background: Endometriosis-associated ovarian cancer (EAOC), primarily comprising ovarian clear cell and endometrioid carcinomas arising in the setting of endometriosis, exhibits substantial molecular heterogeneity. However, the organization of ferroptosis-related transcriptional features across EAOC-associated pathological states remains poorly defined. This study aimed to characterize the heterogeneity of ferroptosis-related transcriptional programs across these pathological states. Methods: A cross-sectional multi-cohort transcriptomic framework was established, comprising one discovery cohort and three independent supportive cohorts. In GSE157153, 1809 detectable FerrDb-derived ferroptosis-related genes were evaluated using global scoring, gene-level screening, state-pattern-based clustering, functional and pathway analyses, and network-based prioritization. Exploratory machine-learning analyses, independent transcriptomic cohorts, and Human Protein Atlas immunohistochemistry were used for complementary characterization and supportive evaluation. Results: The global ferroptosis-related transcriptional score showed only modest variation across pathological states, whereas 929 state-associated candidate genes were organized into four transcriptional programs with divergent expression patterns and distinct functional contexts. These programs showed differential associations with ferroptosis-related metabolic, redox, and signaling pathways, indicating heterogeneous transcriptional remodeling rather than a uniform molecular shift. Network analysis prioritized seven candidate hub genes—CDK1, ERBB2, KRAS, PIK3CA, PTEN, JAK2, and SRC—distributed across multiple programs. Exploratory machine-learning analyses and external supportive evaluations provided heterogeneous, context-dependent evidence for selected candidates. Conclusions: Ferroptosis-related transcriptional alterations across EAOC-associated pathological states are better represented by a heterogeneous program-level architecture than by a single global measure. These findings provide a transcriptomic framework for future mechanistic investigation without establishing ferroptosis-specific functional activity.