Synergic Integration of the miRNome, Machine Learning and Bioinformatics for the Identification of Potential Disease-Modifying Agents in Obstructive Sleep Apnea.

Abstract

Introduction: Understanding the diverse pathogenetic pathways in obstructive sleep apnea (OSA) is crucial for improving outcomes. microRNA (miRNA) profiling is a promising strategy for elucidating these mechanisms.

Objective: To characterize the pathogenetic pathways linked to OSA through the integration of miRNA profiles, machine learning (ML) and bioinformatics.

Methods: This multicenter study involved 525 patients with suspected OSA who underwent polysomnography. Plasma miRNAs were quantified via RNA sequencing in the discovery phase, with validation in two subsequent phases using RT-qPCR. Supervised ML feature selection methods and comprehensive bioinformatic analyses were employed. The associations among miRNA targets, OSA and OSA treatment were further explored using publicly available external datasets.

Results: Following the discovery and technical validation phases in a subset of patients with and without confirmed OSA (n=53), eleven miRNAs were identified as candidates for the subsequent feature selection process. These miRNAs were then quantified in the remaining population (n=472). Feature selection methods revealed that the miRNAs let-7d-5p, miR-15a-5p and miR-107 were the most informative of OSA. The predominant mechanisms linked to these miRNAs were closely related to cellular events such as cell death, cell differentiation, extracellular remodeling, autophagy and metabolism. One target of let-7d-5p and miR-15a-5p, the TFDP2 gene, exhibited significant differences in gene expression between subjects with and without OSA across three independent databases.

Conclusion: Our study identified three plasma miRNAs that, in conjunction with their target genes, provide new insights into OSA pathogenesis and reveal novel regulators and potential drug targets.