Clear cell renal cell carcinoma (ccRCC) is the most common type of renal cell carcinoma and is associated with high morbidity and poor prognosis. Micro-RNAS (miRNAs) have emerged as promising biomarkers for cancer diagnosis and prognosis due to their involvement in cancer progression and development (Ghafouri-Fard et al., 2020). The integration of big omics data from GEO and TCGA, along with data mining and machine learning, has revolutionized the identification of reliable diagnostic and prognostic signatures for various types of cancer. The present study aims to identify a diagnostic and prognostic signature for ccRCC using miRNA data from microarray and NGS experiments in GEO and TCGA.

Differentially expressed miRNAs (DEmiRs) in ccRCC samples compared to normal renal tissue were identified using GEO2R packages in R, with adjusted P<0.05 and log2FC> 1.5 as cutoff criteria. The overlapping DEmiRs were identified, and the target genes of these DEmiRs with strong experimental validation were obtained using miRTargetLink 2.0 (Kern et al., 2021), and the pathway enrichment analysis was performed using ClusterProfiler package in R with KEGG annotation database (Wu et al., 2021). Kaplan-Maier (KM) survival analysis was performed to correlate the survival of patients with higher or lower expression of the identified miRNAs (Lánczky & GyÅ‘rffy, 2021). A support vector machine model was trained and cross-validated to classify tumor samples from matched solid normal tissue samples.

Six datasets, namely GSE11016, GSE12105, GSE47582, GSE73342, GSE151423, and a dataset from TCGA were chosen for the analysis. Results revealed that 14 DEmiRs were consistently differentially expressed in RCC tissues in the microarray datasets, and 26 DEmiRs in the NGS datasets. We identified 9 mRNAs that exhibited a consistent expression trend across all datasets included in the study. We identified 637 genes as targets of the miRNAs under investigation. Pathway enrichment analysis demonstrated that these target genes were significantly enriched in several crucial pathways, including but not limited to AGE-RAGE signaling, MAPK signaling, cellular senescence, toll-like receptor signaling, TNF signaling, PD-L1 expression, and PD-1 checkpoint pathways. Survival analysis revealed that among the 9 signature miRNAs, higher expression of 4 and lower expression of 5 miRNAs were significantly associated with poor survival. Based on these findings, we hypothesize that these identified key miRNAs have the potential to serve as prognostic biomarkers for patients with ccRCC. Using the nine miRNAs identified earlier as features, we trained a support vector machine model on the TCGA dataset. The results of the 10-fold cross-validation demonstrated a high accuracy of 99.23+-0.89% and an AUC of 0.99+-0.007. These findings suggest that the model can accurately and reliably classify tumor samples from normal solid tissue samples.

In summary, this study has identified a nine-miRNA signature that is associated with poor survival outcomes in patients with ccRCC. Moreover, our machine learning model, based on this signature, is capable of distinguishing between tumors and normal tissue samples. Further validation of this model in a clinical cohort would aid in translating our findings into clinical practice, potentially leading to earlier detection and improved follow-up care for ccRCC patients.