Single-Cell-Derived Malignant Epithelial Programs Define Prognostic Risk and Therapeutic Vulnerability in Ovarian Cancer.

Background: Ovarian cancer (OV) is characterized by pronounced intratumoral heterogeneity and unfavorable clinical outcomes, yet clinically applicable biomarkers for risk stratification remain limited. Although bulk transcriptomic models have been widely proposed, they often overlook the cellular origins of prognostic signals. Advances in single-cell RNA sequencing (scRNA-seq) provide an opportunity to resolve malignant epithelial diversity and identify biologically meaningful predictors. Methods: Single-cell transcriptomic datasets were integrated to construct a comprehensive epithelial landscape of OV. Copy number variation profiles inferred by inferCNV were used to distinguish malignant epithelial cells, followed by subpopulation identification and trajectory inference using Slingshot and Monocle3. Subcluster-specific gene signatures were projected onto bulk RNA-seq cohorts, and prognostic genes were screened using Cox regression and LASSO modeling to establish a multigene risk score. The model was validated across multiple independent cohorts. Multi-omics analyses, including pathway enrichment, mutational profiling, tumor mutation burden, immune features, and drug sensitivity prediction, were performed to explore biological and clinical relevance. Functional assays were conducted to validate the role of the key gene PSMB1. Results: Malignant epithelial cells exhibited substantial transcriptional heterogeneity and distinct evolutionary trajectories, revealing subpopulations associated with differential clinical outcomes. A nine-gene risk model derived from these subpopulations demonstrated robust and consistent prognostic performance across multiple datasets. The risk score was closely associated with tumor progression pathways, immune suppression, genomic instability, and reduced therapeutic sensitivity. Notably, PSMB1 was identified as a critical oncogenic factor, and its knockdown significantly inhibited proliferation and colony formation in OV cells. Conclusions: This study establishes a single-cell-informed prognostic framework that links malignant epithelial heterogeneity to clinical outcomes and therapeutic response in OV. The proposed model provides biologically interpretable risk stratification and highlights PSMB1 as a potential therapeutic target, offering new insights for precision oncology.