Identification of novel diagnostic and prognostic microRNAs in sarcoma on TCGA dataset: bioinformatics and machine learning approach

Abstract

Abstract The discovery of unique microRNA (miR) patterns and their corresponding genes in sarcoma patients indicates their involvement in cancer development and suggests their potential use in medical management. MiRs were identified from The Cancer Genome Atlas (TCGA) dataset, with a Deep Neural Network (DNN) employed for novel miR identification. MiRDB facilitated target predictions. Functional enrichment analysis, identify critical pathways, protein-protein interaction network, and diseases/clinical data correlations were explored. COX regression, Kaplan-Meier analyses, and CombioROC was also utilized. The population consisted of 119 females and 142 males, and 1046 miRs were uncovered. Ten miRs was selected for further analysis using DNN. Upon analyzing for gene ontology, it was found that these genes showed enrichment in various activities. We identified a significant association between the overall survival rate of sarcoma patients and miRs levels. The combination of miR.3688 and miR.3936 achieved the greatest diagnostic standing. MiRs have the capability to screen sarcoma patients to identify undetected tumors, predict prognosis, and pinpoint prospective targets for treatment. Further large clinical trials are required to validate our findings.

Keywords

• Artificial intelligence
• Biomarker
• Cancer
• Noncoding RNA
• Prognosis