Comparing classification performance of several types of significant genes to identify key genes in uremia.

OBJECTIVE End-stage renal failure has profound changes in human gene expressions, but the molecular causation of these pleomorphic effects termed uremia is poorly understood. The purpose of this study was to explore key genes in uremia by comparing classification performance of five kinds of significant genes based on the support vector machines (SVM) model. MATERIALS AND METHODS The five kinds of genes were differentially expressed genes (DEGs), differential pathway genes (DPGs), common differential genes between DEGs and DPGs (CDGs), hub genes (HUGs) and common genes of hub genes and DEGs (CHDGs). In detailed, DEGs were detected by linear models for microarray data (Limma) package. Attract method was utilized to capture DPGs from differential pathways. HUGs were determined according to topological centrality analysis of mutual information network (MIN). Subsequently, SVM model was implemented to assess the classification performance of DEGs, DPGs, CDGs, HUGs and CHDGs, depending on its induces the area under the receiver operating characteristics curve (AUC), true negative rate (TNR), true positive rate (TPR) and the Matthews coefficient correlation classification (MCC). RESULTS A total of 166 DEGs, 597 DPGs, 13 CDGs, 29 HUGs and 10 CHDGs were obtained in uremia. By assessing the SVM model classification analysis, CHDGs had the best performance of all with AUC = 0.99, TNR = 1.00, TPR = 0.97 and MCC = 0.95. Hence, we considered the CHDGs as key genes in uremia. CONCLUSIONS Key genes concluded in this investigation might provide vital insights into uremia progression and new therapies.