Establishment of prognostic prediction model based on lipid metabolism related genes in esophageal squamous cell carcinoma by machine learning algorithms

BACKGROUND: Esophageal squamous cell carcinoma (ESCC) is a major cause of cancer-related mortality worldwide, with a high prevalence and poor prognosis in specific regions. Despite advancements in clinical care, the identification of reliable biomarkers for accurate survival prediction remains a significant challenge, hindering personalized treatment. METHODS: This study utilized genomic and clinical data from TCGA and GEO databases, applying ten machine learning algorithms to develop a prognostic model based on lipid metabolism-related genes. The Random Survival Forest (RSF) algorithm was identified as the optimal framework, supplemented by immune infiltration analysis and in vitro functional validation of the key gene ACOT9. RESULTS: The resulting 33-gene signature achieved superior performance, with a C-index of 0.708 and survival prediction area under the curve (AUC) values approaching 1.0 in the training cohort. High-risk scores were significantly correlated with advanced histological grade, tumor stage, and increased regulatory T-cell infiltration. Furthermore, silencing of ACOT9 markedly suppressed the proliferation, migration, and invasion of ESCC cells. CONCLUSIONS: This study provides a robust framework for ESCC risk stratification and identifies ACOT9 as a critical oncogenic driver, offering a novel scientific basis for optimizing clinical management strategies.