A five-gene risk model based on carnitine-related genes for predicting prognosis in ovarian cancer

Abstract Background: Ovarian cancer (OV) remains a highly lethal gynecological malignancy and is associated with unfavorable long-term outcomes. Carnitine participates in metabolic adaptation of tumor cells by modulating fatty acid oxidation and redox balance. This study sought to identify carnitine-associated prognostic biomarkers in OV and to characterize the molecular features linked to these markers. Methods: Publicly available datasets were used throughout the study. Candidate biomarkers were progressively selected through weighted gene co-expression network analysis (WGCNA), differential-expression analysis, regression-based screening, and machine-learning procedures. The selected genes were incorporated into a prognostic risk model, after which patients with OV were assigned to a high-risk group (HRG) or low-risk group (LRG) according to their calculated risk scores. Functional enrichment, immune-infiltration, and drug-sensitivity analyses were subsequently conducted to characterize biological differences associated with risk status. Biomarker expression was further examined using single-cell transcriptomic data to prioritize relevant cell populations. Results: The multistage selection procedure yielded five biomarkers: ALPL, RAB38, SLC4A8, ZMAT4, and IGLON5. Using these genes, the risk model separated patients into the HRG and LRG, with inferior survival observed in the HRG. Across the five biomarkers, 14 shared signaling pathways were identified, among them "Apoptosis", "Hedgehog signaling pathway", and "Ribosome". Scores for 21 immune-cell populations also differed significantly between the two risk categories. Drug-response prediction identified significant between-group differences in the half-maximal inhibitory concentration values of 73 compounds; several agents, including GSK1904529A and ML323, had lower predicted IC₅₀ values in the HRG. Single-cell analysis resolved five cell types and prioritized epithelial cells as the candidate key population. Conclusion: The five genes ALPL, RAB38, SLC4A8, ZMAT4, and IGLON5 were identified as candidate prognostic biomarkers for OV. Their expression was linked to carnitine-associated metabolic characteristics and immune features, supporting further experimental investigation and prospective evaluation of their prognostic value.