Machine learning visceral obesity-related genes based on MARVELD1 model for predicting the prognosis and immune cell infiltration of ovarian cancer

Abstract Background: Visceral obesity elevates high-grade serous carcinoma (HGSOC) risk and prognosis but its effect on biology is underexplored. This study examines obesity-related genes (VORGs) in HGSOC. Methods Using consensus clustering, we analyzed TCGA and GSE53963 datasets with VORGs, uncovering unique immune cell patterns and gene functions. Our prognostic model integrates OS, tumor immunity, ICI response, and chemotherapy outcomes. Single-cell data enrich our understanding of gene expression, particularly highlighting MARVELD1 in tumor stem cells. Experimental validation comprised CCK-8, transwell, and colony formation assays. Results Based on the expression profiles of VORGs associated with OS, HGSOC patients were stratified into four distinct molecular subtypes. The investigation entailed an exploration of the interplay between each molecular subtype, patient prognosis, and immune infiltration dynamics. Subsequently, a robust prognostic prediction model, encompassing four genes (CXCL9, IGF2, FCGBP, and MARVELD1), was meticulously developed. The model utilized differential genes linked to prognosis through molecular subtyping. It categorized patients into high- and low-risk groups, where the high-risk group exhibited significantly worse outcomes (P < 0.001). Elevated scores were associated with increased stem cell indices, reduced mutation burdens, and heightened immunosuppression. Moreover, these scores showed significant correlations with immune checkpoint genes and chemotherapy sensitivity, emphasizing their crucial role in predicting treatment responses. Turning our attention to finer granularity, analyzing single-cell data revealed diverse gene expression among the four models, especially MARVELD1's prominence in tumor stem cells. MARVELD1 impacted early cellular functions through pathways like reactive oxygen species. Lastly, to validate our findings, in vitro experiments provided compelling evidence that interfering with MARVELD1 effectively curbed the proliferation, invasion, and clone formation potential of ovarian cancer cells. Conclusion This study examines VORGs' impact on HGSOC biology, using scRNA-seq and bulk RNA-seq data, presenting a novel risk model correlated with prognosis, immune microenvironment, and drug effectiveness.