[Screening of characteristic genes of salivary gland adenoid cystic carcinoma based on weighted co-expression network and machine learning].

Abstract

Purpose: To identify potential biomarkers of salivary gland adenoid cystic carcinoma to further understand the potential pathogenesis of adenoid cystic carcinoma.

Methods: Two microarray datasets (GSE59701, GSE88804) were downloaded from NCBI GEO database. LIMMA software package was used to screen SACC differentially expressed genes. WGCNAs were used to find the important module genes that were most associated with SACC. Two machine learning methods(LASSO and SVM-RFE) were used to identify Hub genes. Subsequently, ROC curve used to predict SACC was developed to determine the diagnostic effect. R4.2.1 software was used for statistical analysis.

Results: Three hub genes(GABBR1, EN1 and LINC01296) were identified, and a ROC curve with high predictive performance (AUC, 1.000-1.000) was established.

Conclusions: Three hub genes(GABBR1, EN1 and LINC01296) were obtained by WGCNA, LASSO, SVM-RFE as potential biomarkers of SACC, and the findings of this study provide a foothold for future research on potential key genes of SACC, and a target basis for the early diagnosis and treatment of SACC.