The identification of metabolomic biomarkers relies on the analysis of large cohorts of patients compared to healthy controls followed by the validation of markers in an independent sample set. Indeed, circulating biomarkers should be causally linked to pathology to ensure that changes in the marker precede changes in the disease. However, this approach becomes unfeasible in rare diseases due to the paucity of samples, necessitating the development of new methods for biomarker identification. The present study describes a novel approach that combines samples from both mouse models and human patients to identify biomarkers of OPMD. We initially identified a pathology-specific metabolic fingerprint in murine dystrophic muscle. This metabolic fingerprint was then translated into (paired) murine serum samples and then to human plasma samples. This study identified a panel of nine candidate biomarkers that could predict muscle pathology with a sensitivity of 74.3% and specificity of 100% in a random forest model. These findings demonstrate that the proposed approach can identify biomarkers with good predictive performance and a higher degree of confidence in their relevance to pathology than markers identified in a small cohort of human samples alone. Therefore, this approach has a high potential utility for identifying circulating biomarkers in rare diseases.
Keywords: LC-MS; biomarker; metabolomics; oculopharyngeal muscular dystrophy; random forest.