miRNA biomarkers-based diagnosis of diffuse large B-cell lymphoma using machine learning

Background: Traditional diagnostic methods like excisional biopsy, fine needle aspiration (FNA), and core needle biopsy (CNB) are often challenged by sampling errors and high false-negative rates. Our research shifts focus to microRNA (miRNA) expression profiling, leveraging the stability of miRNA molecules and advanced RNA extraction methods. Although the oncogenic potential of miRNAs in B-cell lymphoma has been studied since 2005, and various dysregulated miRNAs in diffuse large B-cell lymphoma (DLBCL) patients have been reported in the scientific literature, there has been limited research investigating these miRNAs using ML algorithms. Methods: This study presents an innovative approach to the diagnosis of DLBCL using a machine-learning (ML) system based on miRNA analysis. We first identified 54 miRNAs associated with DLBCL, combining them with 54 random miRNAs to create a training dataset for ML classifiers. This dataset was processed using various ML classifiers through the Waikato Environment for Knowledge Analysis (WEKA) software. In addition to miRNA profiling, our study also explored the biological pathways associated with these miRNAs using the Database for Annotation, Visualization, and Integrated Discovery (DAVID) and Kyoto Encyclopedia of Genes and Genomes (KEGG) databases. Results: Our training model achieved a notable accuracy of 93.52%. The performance was further validated with three independent datasets derived from actual tumor samples, showing best accuracies from 86.36% to 100%. We identified several enriched pathways, such as the PI3K and FoxO signaling pathways, that are significantly implicated in DLBCL. These findings not only validate known associations but also reveal potential new avenues for understanding DLBCL pathogenesis. Conclusions: Our paper demonstrates that ML-assisted miRNA analysis can serve not only as a diagnostic tool for the onset of DLBCL but also as a discovery tool to predict specific genes, pathways, and sequence motifs as targets for further investigation.