A FIVE-GENE SUPPORT VECTOR MACHINE CLASSIFIER TO PREDICT TWO POTENTIAL PRIMARY SITES OF METASTATIC CERVICAL CARCINOMA FROM UNKNOWN PRIMARY

PURPOSE: Metastatic cervical carcinoma from unknown primary (MCCUP) accounts for 1–4% of all head and neck tumors, and identifying the primary site in MCCUP is challenging. As two type of potential primary site of MCCUP, esophageal squamous cell carcinoma (ESCC) has a strong resemblance to head and neck squamous cell carcinoma (HNSCC). The aim of the current study was to find a novel and effective method to determine the two types of potential primary site of MCCUP. METHODS: The gene expression profiles of ESCC and HNSCC from Gene Expression Omnibus (GEO) were extracted and differentially expressed genes (DEGs) were identified. The common and difference set between HNSCC’s and ESSC’s common DEGs were subjected to gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis and protein-protein interaction(PPI) network analysis. Several machine learning methodologies (k-Nearest Neighbor (KNN), Random Forest(RF), support vector machine (SVM)) were adopted to construct the model to discriminate the two types of carcinoma. RESULTS: The two types of squamous cell carcinoma (SCC) shared some common characteristics as well as differences. A predictive SVM model consisting of a five-gene signature was established that could effectively discriminate between the two types of carcinoma. CONCLUSIONS: A five-gene (HSPD1, PCP4, CXCL11, GMNN, and PITX1) predictive SVM model might has clinical utility for the accurate diagnosis of MCCUP. However, large-scale and comprehensive research is needed to clarify our results. CLINICAL IMPLICATIONS: It might have clinical utility for the accurate diagnosis of MCCUP and provide useful guidance for personalized and precision therapy.