Exploring and Validating Prognostic Biomarkers Related to Sphingolipid Metabolism in Gastric Cancer through Machine Learning.

Background: Sphingolipid metabolism (SM) has been implicated in the progression of gastric cancer (GC). However, its potential as a prognostic biomarker in GC remains underexplored. This study investigates the feasibility of using SM to predict GC prognosis.

Methods: Sphingolipid metabolism-related genes (SMRGs) were extracted from the GeneCards database, and differentially expressed genes (DEGs) were identified using the TCGA-STAD and GSE84437 gastric cancer datasets. Univariate Cox regression analysis was performed to identify genes associated with survival. Lasso-Cox and random survival forest analyses were employed to identify key survival-related genes, followed by multivariate Cox regression to establish a prognostic model and calculate the sphingolipid metabolism score (SMscore). The lasso-Cox analysis further assessed the prognostic significance of clinical traits and the SMscore. Hyperparameters were optimized using machine learning models to achieve the most accurate prognostic model. The potential utility of the SMscore in GC prognosis was evaluated, and hub gene expression was validated through immunohistochemistry (IHC) staining.

Results: ELOVL4, NOS3, and ABCA2 were identified as key prognostic genes from a pool of 556 SMRGs. The optimal prognostic model was developed and validated, demonstrating robust predictive performance. IHC staining revealed increased expression of ELOVL4 and NOS3 in tumor tissues, which correlated significantly with poor prognosis.

Conclusion: Bioinformatics analysis and IHC validation suggest that ELOVL4 expression may serve as a prognostic biomarker for GC, providing new insights for prognosis prediction and therapeutic target development in gastric cancer.