Gene expression profiles reveal key genes for early diagnosis and treatment of adamantinomatous craniopharyngioma. (October 2018)

• Record Type: Journal Article
• Authors: Yang, Jun; Hou, Ziming; Wang, Changjiang; Wang, Hao; Zhang, Hongbing
• Is Part Of: Cancer gene therapy. Volume 25:Number 9/10(2018)
• Page Start: 227
• Page End: 239
• Publication Date: 2018-10
• Subjects: Cancer -- Gene therapy -- Periodicals; Gene therapy -- Periodicals; Gene Therapy -- Periodicals; Neoplasms -- therapy -- Periodicals; Electronic journals; 616.99406
• DOI: 10.1038/s41417-018-0015-4
• Languages: English
• ISSNs: 0929-1903

Abstract

Abstract Adamantinomatous craniopharyngioma (ACP) is an aggressive brain tumor that occurs predominantly in the pediatric population. Conventional diagnosis method and standard therapy cannot treat ACPs effectively. In this paper, we aimed to identify key genes for ACP early diagnosis and treatment. Datasets GSE94349 and GSE68015 were obtained from Gene Expression Omnibus database. Consensus clustering was applied to discover the gene clusters in the expression data of GSE94349 and functional enrichment analysis was performed on gene set in each cluster. The protein–protein interaction (PPI) network was built by the Search Tool for the Retrieval of Interacting Genes, and hubs were selected. Support vector machine (SVM) model was built based on the signature genes identified from enrichment analysis and PPI network. Dataset GSE94349 was used for training and testing, and GSE68015 was used for validation. Besides, RT-qPCR analysis was performed to analyze the expression of signature genes in ACP samples compared with normal controls. Seven gene clusters were discovered in the differentially expressed genes identified from GSE94349 dataset. Enrichment analysis of each cluster identified 25 pathways that highly associated with ACP. PPI network was built and 46 hubs were determined. Twenty-five pathway-related genes that overlapped with the hubs in PPI network were used as signatures to establish the SVM diagnosis model for ACP. The prediction accuracy of SVM model for training, testing, and validation data were 94, 85, and 74%, respectively. The expression ofCDH1, CCL2, ITGA2, COL8A1, COL6A2, andCOL6A3 were significantly upregulated in ACP tumor samples, whileCAMK2A, RIMS1, NEFL, SYT1, andSTX1A were significantly downregulated, which were consistent with the differentially expressed gene analysis. SVM model is a promising classification tool for screening and early diagnosis of ACP. The ACP-related pathways and signature genes will advance our knowledge of ACP pathogenesis and benefit the therapy improvement.