Predictive Ability of Blood Plasma Carnitines in Cardiovascular Diseases
Zsuzsanna Ament, Investigator Scientist, MRC Human Nutrition Research
Hypertrophic cardiomyopathy (HCM) is the most common inherited heart disease often caused due to single gene mutations, principally encoding sarcomeric components. This results in abnormal thickening of heart muscle and predisposes patients to malignant arrhythmias and sudden death. Current therapies are based on empirical evidence and are often inadequate. In the current study, metabolic alterations in patients with HCM were compared to controls and to patients with aortic stenosis (AS) or ischaemia. To date, the majority of similar studies examining cardiac disease are confined to the use of peripheral blood as a surrogate for the myocardium. However, there is a potential for confounding metabolic signals from other organs. In order to mitigate these effects in the present study, samples were collected directly from the coronary sinus (CS) and the aortic root (AR), and compared with the femoral vein (FV) as a measure of systemic metabolism. Healthy heart muscles generate approximately 90% of ATP from ß-oxidation requiring fatty acid transport to the mitochondria, alongside adequate CPT-1 and 2 activity and adequate amounts of free carnitine to shuttle the fatty acids into the mitochondria. In order to measure the rate of fatty acid oxidation and the levels of free carnitine, a targeted metabolomic experiment was carried out measuring 38 carnitines by LC-MS/MS. We sought to determine carnitines which differentiate the patient groups. Partial least squares-discriminant analysis (PLS-DA) and random forest (RF) methods were applied for predictive modelling. The best models were found comparing HCM vs. AS, and control vs. AS groups, suggesting that the AS group has the biggest impact on carnitine metabolism. In both cases, the discriminating carnitines were established by RF and the number of selected carnitines were optimised based on classification error. Selected carnitines were then used to build multivariate receiver operating characteristic (ROC) curves resultin
|
|