Evaluation of pyroptosis-associated genes in endometrial cancer based on the 101- combination machine learning framework and multi-omics data

Background Endometrial cancer (EC) represents a common malignancy within gynecological cancers, characterized by a notably high mortality rate. The absence of reliable prognostic biomarkers significantly impairs the effectiveness of predictive, preventive, and personalized medicine (PPPM/3PM) strategies. Pyroptosis, a distinct form of programmed cell death, has been closely linked to anti-cancer immune responses. Nonetheless, the precise role of pyroptosis in the context of EC remains elusive. Methods Pyroptosis-associated genes (PAGs) were screened in Msigdb. We used consensus clustering to classify PAGs from TCGA-UCEC into two clusters, and examined their characteristics. The Seurat package was employed to analyze significant PAGs in EC single-cell data. The mime package was utilized to screen suitable machine learning approaches and build models. A nomogram was constructed to validate the model's performance. Additionally, CIBERSORT was used to evaluate immune infiltration results, and TIDE scores from the TCIA database were applied to assess EC patients' responses to immune checkpoint therapy. Subsequently, we performed PAG-related pathway analysis in EC patients with or without response to PD-1 therapy using the CellChat module in Seurat. Finally, the OncoPredict package was used to predict drug sensitivity in EC patients. Results A consensus PAGs ("CASP3", "CHMP3", "CYCS", "GSDMD", "IRF1", and "NOD1") was constructed based on a 101-combination machine learning computational framework, demonstrating outstanding performance in predicting prognosis and clinical translation. We observed distinct biological functions and immune cell infiltration in the tumor microenvironment between the high- and low-risk groups. Notably, the immunophenoscore (IPS) score showed a significant difference between risk subgroups, suggesting a negative response to PD-1 in the high-risk group. Potential drugs targeting specific risk subgroups were also identified. Conclusion Our study constructed an PAGs that can serve as a promising tool for prognosis prediction, targeted prevention, and personalized medicine in EC.