Bioinformatics Analysis Identifies Lipid Droplet-Associated Gene Signatures as Promising Prognostic and Diagnostic Models for Endometrial Cancer

Abstract

Background

Effective diagnostic and prognostic tools are critical for early detection and improved outcomes in endometrial cancer (EC). Although metabolic dysregulation plays a key role in EC pathogenesis, the clinical relevance of lipid droplet–associated genes (LDAGs) remains largely unexplored. This study aims to establish LDAG-based gene signatures with strong diagnostic and prognostic potential in EC.

Aims

To identify LDAG signatures with prognostic and diagnostic utility in EC.

Methods and Results

A curated set of LDAGs was systematically analyzed across publicly available EC datasets to identify differentially expressed LDAGs (DE-LDAGs). Survival-associated DE-LDAGs were then identified using univariate Cox regression. A four-gene prognostic model was developed through LASSO-based feature selection followed by multivariate Cox regression and validated using Kaplan–Meier survival and time-dependent receiver operating characteristic (ROC) analyses. From the same pool of survival-associated DE-LDAGs, a six-gene diagnostic model was constructed using LASSO, ROC analysis, and logistic regression. Model performance was evaluated using ROC curves and support vector machine (SVM) classification. Functional enrichment and protein–protein interaction (PPI) network analyses were conducted to assess the biological relevance of the identified genes.

Our results demonstrate that the four-gene prognostic model (LMLN, LMO3, PRKAA2, and RAB10) stratified EC patients into high- and low-risk groups with significantly different survival outcomes (p < 0.05; time-dependent AUC > 0.70). The six-gene diagnostic model (AIFM2, ABCG1, LIPG, DGAT2, LPCAT1, and VCP) demonstrated near-perfect classification of tumor versus normal tissues (AUC ≈0.99 in ROC analysis; 99.8% accuracy in SVM analysis). Functional enrichment linked DE-LDAGs to lipid metabolism, ER stress response, cholesterol homeostasis, and autophagy, underscoring their biological relevance in EC pathobiology.

Conclusion

This study provides the first comprehensive analysis of LDAGs in EC, establishing robust prognostic and diagnostic gene signatures with strong biological relevance. These signatures support a metabolism-driven framework for EC classification and may offer potential clinical utility in early detection, risk stratification, and personalized treatment.