Metabolic profiling of Parkinson's disease: evidence of biomarker from gene expression analysis and rapid neural network detection

Abstract

Background: Parkinson's disease (PD) is a neurodegenerative disorder. The diagnosis of Parkinsonism is challenging because currently none of the clinical tests have been proven to help in diagnosis. PD may produce characteristic perturbations in the metabolome and such variations can be used as the marker for detection of disease. To test this hypothesis, we used proton NMR and multivariate analysis followed by neural network pattern detection.

Methods & results: 1H nuclear magnetic resonance spectroscopy analysis was carried out on plasma samples of 37 healthy controls and 43 drug-naive patients with PD. Focus on 22 targeted metabolites, 17 were decreased and 5 were elevated in PD patients (p < 0.05). Partial least squares discriminant analysis (PLS-DA) showed that pyruvate is the key metabolite, which contributes to the separation of PD from control samples. Furthermore, gene expression analysis shows significant (p < 0.05) change in expression of PDHB and NPFF genes leading to increased pyruvate concentration in blood plasma. Moreover, the implementation of 1H- NMR spectral pattern in neural network algorithm shows 97.14% accuracy in the detection of disease progression.

Conclusion: The results increase the prospect of a robust molecular definition in detection of PD through the early symptomatic phase of the disease. This is an ultimate opening for therapeutic intervention. If validated in a genuinely prospective fashion in larger samples, the biomarker trajectories described here will go a long way to facilitate the development of useful therapies. Moreover, implementation of neural network will be a breakthrough in clinical screening and rapid detection of PD.

Figure 1

Heat map differentiation of metabolite. Average metabolite variability of blood plasma between PD patients (n = 43) and healthy controls (n = 37) are shown. Cluster analyses of the 22 differentially altered metabolites are selected based on significance P value (P < 0.05). The heat map depicts high (red) and low (green) relative levels of metabolite variation.

Figure 2

Partial least square discriminant analysis. PLS-DA scores plot showing a significant separation between control subjects (n = 37) and unmedicated PD patients (n = 43) using complete digital maps. The observations coded according to class membership: black square = controls; red square = PD patients.

Figure 3

Biomarker detection. The loading coefficient map showing that myoinositol, glucitol, citrate, acetate, succinate and pyruvate were predominantly responsible for the classification of groups.

Figure 4

Pyruvate dehydrogenase component interacting genes. The systems biological approach showing the complex interaction of pyruvate dehydrogenase components (red) with 40 genes.

Figure 5

Distribution of patients and healthy individuals. Neural network classification of disease for 35 individuals (x-axis) with known clinical information. Values (y-axis) are predicted value over the trained network and are 0 to3; values ≥ 0.2 reflect a neural-network classification of "normal," and values ≤ 0.2 reflect a neural-network classification of "PD". Individuals denoted by a Violet rhombus are normal, red arrow are clinically stage 1 PD, blackened circle are stage 2, blackened square are stage 3 PD respectively.