Construction of machine learning-based prognostic model of centrosome amplification-related genes for esophageal squamous cell carcinoma

To construct a prognostic model of centrosome amplification-related genes (CARGs) by machine learning and evaluate its prediction performance for the prognosis of esophageal squamous cell carcinoma (ESCC).

CARGs were obtained from Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) datasets. The RNA-sequencing (RNA-seq) transcriptome datasets of ESCC were retrieved from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO), and assigned into training and validation sets, respectively. Subsequently, single-sample gene set enrichment analysis (ssGSEA) and weighted gene co-expression network analysis (WGCNA) were employed to screen CARGs. A prognostic model of CARGs was constructed by incorporating 12 machine learning algorithms, and univariate and multivariate Cox regression analyses were applied to evaluate whether the 12 models as an independent prognostic factor or not. Eventually, 15 paired ESCC and adjacent non-tumor tissue samples were collected from the Department of Gastroenterology of the Second Affiliated Hospital of Army Medical University, and real-time quantitative PCR (RT-qPCR) and immunohistochemistry staining were performed to detect the expression of these genes ESCC samples.

Our 9-CARGs prediction model for ESCC prognosis was constructed. RT-qPCR confirmed that the mRNA expression levels of DENR, TRIP13, BRCA2, TTF2, TCFL5 and NUP188 were significantly higher in ESCC tissues than normal tissues (P<0.05), and the protein levels of DENR, TRIP13, TTF2 and TCFL5 were also elevated when compared to normal tissues.

DENR, TRIP13, TTF2 and TCFL5 are highly expressed and closely associated with poor prognosis of ESCC, suggesting their potential roles in the pathogenesis and progression of this malignancy.