Metabolomics focuses on the study of metabolic changes occurring in varioussystems and utilizes quantitative and semi-quantitative measurements of multiple metabolites in parallel. Mass spectrometry (MS) is the most ubiquitous platform in this field, as it provides superior sensitivity regarding measurements of complex metabolic profiles in biological systems. When combined with MS, multivariate statistics and advanced machine learning algorithms provide myriad opportunities for bioinformatics insights beyond simple univariate data comparisons. In this dissertation, the application of MS-based metabolomics is introduced with an emphasis on biomarker discovery for human disease detection. To advance disease diagnosis using MS-based metabolomics, numerous statistical techniques have been implemented in this research including principal component analysis, factor analysis, partial least squares-discriminant analysis (PLS-DA), orthogonal PLS-DA, random forest, receiver operating characteristic analysis, as well as functional pathway/enzyme enrichment analyses. These approaches are highly useful for improving classification sensitivity and specificity related to disease-induced biological variation and can help identify useful biomarkers and potential therapeutic targets. It is also shown that MS-based metabolomics can distinguish between clinical and prodromal disease as well as similar diseases with related symptoms, which may assist in clinical staging and differential diagnosis, respectively. Additionally, MS-based metabolomics is shown to be promising for the early and accurate detection of diseases, thereby improving patient outcomes, and advancing clinical care. Herein, the application of MS methods and chemometric statistics to the diagnosis of breast cancer, coccidioidomycosis (Valley fever), and senile dementia (Alzheimer's disease) are presented and discussed. In addition to presenting original research, previous efforts in biomarker discovery will be synthesized and appraised. A Comment will be offered regarding the state of the science, specifically addressing the inefficient model of repetitive biomarker discovery and the need for increased translational efforts necessary to consolidate metabolomics findings and formalize purported metabolic markers as laboratory developed tests. Various factors impeding the translational throughput of metabolomics findings will be carefully considered with respect to study design, statistical analysis, and regulation of biomedical diagnostics. Importantly, this dissertation will offer critical insights to advance metabolomics from a scientific field to a practical one including targeted detection, enhanced quantitation, and direct-to-consumer considerations.

Epidemiological studies have identified obesity as a risk factor for numerous chronic diseases such as adult onset diabetes, hypertension, and hypercholesterolemia. In both humans and laboratory animals, high-fat diets have been shown to cause obesity. Increases in dietary fat lead to increased energy consumption and, consequently, significant increases in body fat content. CD36 has been implicated in fat perception, preference, and increased consumption, but it is yet to be tested using a behavior paradigm. To study the effect of CD36 on fat taste transmission and fat consumption, four CD36 knockout (experimental) mice and four Black 6 wildtype (control) mice underwent 20 days of fat preference and perception testing. Both groups of mice were exposed to foods with progressively increasing fat content (10%, 12.5%, 15% 17.5%, 20%, 45%) in order to assess the effect of CD36 on fat preference. Afterward, the mice were subjected to an aversive conditioning protocol designed to test the effect of CD36 on fat taste perception; development of a conditioned taste aversion was indicative of ability to taste fat. Especially, knockout mice exhibited diminished preference for and reduced consumption of fat during preference testing and were unable to identify fat taste as the conditioned stimulus during aversive conditioning. A repeated measures ANOVA with Bonferroni correction revealed a significant main effect of group on fat consumption, energy intake, and weight. Linear regression revealed CD36 status to account for a majority of observed variance in fat consumption across both phases of the experiment. These results implicate CD36 in fat taste perception and preference and add to the growing body of evidence suggesting fat as a primary taste.