Year: 2019
Type: dissertation
Study design: Observational Study
Abstract: Asthma and chronic obstructive pulmonary disease (COPD) are chronic inflammatory respiratory illnesses with increasing mortality rates. Despite distinct pathophysiological differences, the differential diagnosis of these diseases can be challenging in some patients. Currently, physicians rely on patients’ history and therapy trials leading to misdiagnoses and disease management errors. Therefore, better diagnostic tests are needed. A recent untargeted proton nuclear magnetic resonance (1H-NMR) metabolomics study revealed 50 urine metabolites as candidate biomarkers for differentiating asthma from COPD. Since clinical validation of candidate biomarkers requires robust assays, we developed liquid chromatography-tandem mass spectrometric (LC-MS/MS) methods to meet this purpose. The metabolites were divided into four subgroups based on structure/concentration. Differential isotope labeling and hydrophilic interaction LC were utilized for method development. Methods were fully validated for clinical testing according to the FDA and European Medicines Agency guidelines. Two strategies of urine normalization were evaluated (creatinine vs. osmolality). Urine samples from asthma (n=51) and COPD (n=78) patients were analyzed for 41 metabolites and statistical analysis was performed using partial least square-discriminant analysis (PLS-DA). The endogenous nature of the metabolites sometimes hindered the direct application of the regulatory validation guidelines. Accordingly, novel approaches were adopted to address the unexpected challenges. For example, since metabolite-free urine is not available, blank surrogate matrix and pooled urine were used during validation. The isotopic interference from the metabolites on their internal standards dictated atypical optimization and validation experiments. PLS-DA successfully separated asthma from COPD using a validated model (R2Y= 0.851, Q2= 0.759). Nineteen metabolites were identified as the most significant biomarkers for disease differentiation. To test the model, 27 samples were run blindly resulting in 82% accuracy. The involved biochemical pathways were identified. To expedite sample analysis, we compared between absolute (fully validated), relative and single point quantification strategies, since the last two are widely used in metabolomics studies. In general, while the three methods were precise, the fully validated method provided the most accurate data. Finally, to test the ruggedness of the analytical platforms, 17 urine samples were reprocessed after 28-months storage. The PLS-DA classification of the reanalyzed samples was 100% identical to their initial results.
Author: Mona Khamis Hamada
Institution: University of Saskatchewan
Language: English
Subfield: Molecular Biology
Domain: Life Sciences
Keywords: asthma, chronic obstructive pulmonary disease, differential diagnosis, urine metabolomics, biomarkers, LC-MS/MS method validation, PLS-DA classification, creatinine, osmolality, isotopic labeling, hydrophilic interaction liquid chromatography, relative quantification, sample storage stability, respiratory diseases
Open Access status: green
OpenAlex ID: W2957591575
