Abstract
The prevalence of papillary renal cell carcinomas (PRCCs) is estimated to be between 10% and 15%. At present, there is no effective therapeutic approach available for patients with advanced PRCCs. The molecular biomarkers associated with PRCC diagnoses have been rarely studied compared to renal clear cell carcinomas; therefore, the necessity for the identification of novel molecular biomarkers to aid in the early identification of this disease. Bioinformatics and artificial intelligence technologies have become increasingly important in the search for diagnostic biomarkers for early cancer detection. In this study, three genes-BCL11A, NTN5, and OGN-were identified as diagnostic biomarkers using the Cancer Genome Atlas (TCGA) database and deep learning techniques. To identify the differential expression genes (DEGs), ribonucleic acid (RNA) expression profiles of PRCC patients were analyzed using a machine learning approach. A number of molecular pathways and coexpressions of DEGs have been analyzed and a correlation between DEGs and clinical data has been determined. Diagnostic markers were then determined via machine learning analysis. The 10 genes selected with the highest variable importance value (more than 0.9) were further investigated, with six upregulated (BCL11A, NTN5, SEL1L3, SKA3, TAPBP, SEMA6A) and four downregulated (OGN, ADCY4, SMOC2, CCL23). A combined receiver operating characteristic (ROC) curve analysis revealed that the BCL11A-NTN5-OGN genes, which have specificity and sensitivity values of 0.968 and 0.901, respectively, can be used as a diagnostic biomarker for PRCC. In general, the genes introduced in this study may be used as diagnostic biomarkers for the early diagnosis of PRCC, thus providing the possibility of early treatment and preventing the progression of the disease.