In the veterinary field, the utility of disease-identification models that use comprehensive circulating microRNA (miRNA) profiles produced through measurements based on next-generation sequencing (NGS) remains unproven. To integrate NGS technology with automated machine learning (autoML) to create a comprehensive circulating miRNA profile and to assess the clinical utility of a disease-screening model derived from this profile. The study involved dogs diagnosed with or being treated for various diseases, including tumors, across multiple veterinary clinics (n = 254), and healthy dogs without apparent diseases (n = 91). miRNA was extracted from EDTA-treated plasma, and a comprehensive analysis was conducted of one million reads per sample using NGS. Then autoML technology was applied to develop a diagnostic model based on miRNA. Among these models, the one with the highest performance was chosen for evaluation. The diagnostic model, based on the comprehensive circulating miRNA profile developed in this study, achieved an AUC score of 0.89, with a sensitivity of 85 % and a specificity of 88 % for the disease samples. The miRNA-based diagnostic model demonstrated high sensitivity for disease groups and has the potential to be an effective screening test. This study indicates that a comprehensive miRNA profile in dog plasma could serve as a highly sensitive blood biomarker.
Keywords: Automated machine learning; Canine; Next-generation sequencing; microRNA.
© 2024 The Authors. Published by Elsevier Ltd.