NMR-linked metabolomics analysis was employed to investigate urinary and human plasma profiles collected from Niemann Pick type C1 disease patients (NP-C1), in addition to aqueous extracts of liver samples of an NP-C1 mouse model. NP-C1 is a lysosomal storage disorder caused by mutations in the lysosomal proteins NPC1 and NPC2, which are involved in lysosomal cholesterol trafficking. NP-C1 disease is a fatal genetic disorder, characterised by neurodegeneration and hepatic damage. Miglustat (MGS) is the only approved drug for this disease, and consequently, plasma and urine samples collected from MGS-treated patients were also investigated.
The ability of 1H NMR analysis to detect a wide range of metabolites simultaneously served to characterize the metabolic profiles of urine, plasma and hepatic tissue samples investigated in order to perform linked multivariate analysis (MVA). Additionally, MGS was identified in urine samples collected from NP-C1 treated patients. MVA employing both parametric and machine learning-based techniques was conducted to classify samples according to their disease status, and also to seek biomarkers that could aid in the diagnosis and/or prognosis of the disease.
Moreover, a new technique was introduced in a metabolomics context, Correlated Component Regression (CCR), and the suitability of Random Forests (RFs) for variable selection was also explored.
We were able to differentiate urine samples collected from NP-C1 patients from those collected from heterozygous controls, and also propose several metabolites as NP-C1 urinary biomarkers such as bile acids, 2-hydroxy-3-methylbutyrate, 3-aminoisobutyrate, 5-aminovalerate, trimethylamine, methanol, creatine and quinolinate. The 1H NMR linked metabolomics study of plasma samples revealed major distinctions among the groups investigated, metabolic alterations ascribable to the disease pathology were mainly observed as changes in the lipoprotein profiles of NP-C1 patients. Hepatic tissue extracts analysed revealed major disturbances in amino acid metabolism, along with impairments in the NAD+/NADH production and redox status.
Gut microbiota and bile acid metabolism were also highlighted as features altered in NP-C1 disease. CCR linked to Linear Discriminant Analysis was evaluated as a new tool for metabolomics analysis, giving accurate results when compared to alternative techniques tested.
Additionally, the suitability of Random Forests and associated recursive feature elimination for variable selection in metabolomics studies was contrasted, suggesting that those strategies relying on a variable ranking to select the top features for discrimination are more suitable for metabolomics investigations than those that iteratively remove a percentage of the least effective features until the classification performance decays.