Abstract
In newborn errors of metabolism, biomarkers are urgently needed for disease screening, diagnosis, and monitoring of therapeutic interventions. This article describes a 2-step approach to discovermetabolic markers, which involves (1) the identification ofmarker candidates and (2) the prioritization of thembased on expert knowledge of diseasemetabolism. For step 1, the authors developed a new algorithm, the biomarker identifier (BMI), to identifymarkers fromquantified diseased versus normal tandemmass spectrometry data sets. BMI produces a ranked list ofmarker candidates and discards irrelevant metabolites based on a quality measure, taking into account the discriminatory performance, discriminatory space, and variance ofmetabolites’ concentrations at the state of disease. To determine the ability of identified markers to classify subjects, the authors compared the discriminatory performance of several machine-learning paradigms and described a retrieval technique that searches and classifies abnormal metabolic profiles from a screening database. Seven inborn errors of metabolism— phenylketonuria (PKU), glutaric acidemia type I (GA-I), 3-methylcrotonylglycinemia deficiency (3-MCCD), methylmalonic acidemia (MMA), propionic acidemia (PA), medium-chain acylCoAdehydrogenase deficiency (MCADD), and 3-OH longchain acyl CoA dehydrogenase deficiency (LCHADD)—were investigated. All primarily prioritized marker candidates could be confirmed by literature. Somenovel secondary candidateswere identified (i.e., C16:1 andC4DCfor PKU, C4DCfor GA-I, and C18:1 forMCADD), which require further validation to confirmtheir biochemical role during health and disease.