Identification of a miRNA signature as a diagnostic and prognostic marker in renal cell carcinoma

Abstract

Clear cell renal cell carcinoma (ccRCC) is the most common subtype of renal cell carcinoma (RCC). If diagnosed in later stages, ccRCC is associated with high renal cancer related morbidity and poor prognosis. Recently, microRNAs (miRNAs) have attracted interest as potential diagnostic and prognostic biomarkers due to their important role in cancer development and progression. Availability of big omics data in the cancer genome atlas (TCGA) coupled with data mining and machine learning have revolutionized the identification of robust diagnostic and prognostic signatures in different types of cancers. In this study, we have utilized the miRNA sequencing data of 516 ccRCC patients from TCGA to identify a diagnostic and prognostic signature by using a combined approach of differential expression analysis, survival analysis and machine learning. Differential expression analysis identified 30 downregulated and 20 upregulated miRNAs in the primary tumor as compared to solid tissue normal samples. Out of these 50 differentially expressed miRNAs, higher expression of 7 and lower expression of 6 miRNAs were found to be significantly associated with poor survival when analyzed using the Kaplan-Maier survival method. Pathway enrichment analyses related to the differentially expressed miRNAs revealed that fatty acid biosynthesis was the most significantly enriched KEGG pathway while proteoglycans in cancer pathway was enriched by the highest number of survival-associated miRNAs target genes. Differential expression and association with poor survival was used as a prefilter for training a support vector machine model capable of classifying tumor samples from solid tissue normal samples with an accuracy and precision of 99.23% and 98.50%, respectively. We have identified here a nine-miRNA signature in ccRCC patients that is capable of segregating tumor from normal tissue samples with high accuracy and precision. The future validation of this classification model in in a clinical cohort will support translation of these findings into clinical practice for early detection and follow-up of ccRCC.