Abstract
Bicuspid aortic valve (BAV) is the most common congenital heart disease. The current study aims to construct a diagnostic model based on metabolic profiling as a non-invasive tool for BAV screening. Blood serum samples were prepared from an estimation group and a validation group, each consisting of 30 BAV patients and 20 healthy individuals, and analyzed by liquid chromatography-mass spectrometry (LC-MS). In total, 2213 metabolites were detected and 41 were considered different. A model for predicting BAV in the estimation group was constructed using the concentration levels of monoglyceride (MG) (18:2) and glycerophospho-N-oleoyl ethanolamine (GNOE). A novel model named Zhongshan (ZS) was developed to amplify the association between BAV and the two metabolites. The area under curve (AUC) of ZS for BAV prediction was 0.900 (0.782-0.967) and was superior to all single-metabolite models when applied to the estimation group. Using optimized cutoff (-0.1634), ZS model had a sensitivity score of 76.7%, specificity score of 90.0%, positive predictive value of 80% and negative predictive value of 85.0% for the validation group. These results support the use of serum-based metabolomics profiling method as a complementary tool for BAV screening in large populations.