In this Master's thesis, we investigated the potential for early detection of anorexia, juvenile idiopathic arthritis (JIA), tics, and endometriosis based on the urinary metabolome. In total, we analysed 489 urine samples obtained from 24 individuals with anorexia, 36 with JIA, 44 with tics, 11 with endometriosis, and 71 healthy controls. The samples were measured using the 1H NMR method, and the spectra were processed with Chenomx software. Following quantification of 338 metabolites, we used the MetaboAnalyst tool to identify metabolic pathways and metabolites that show statistically significant differences between each disease group and healthy individuals. In anorexia, the most pronounced changes were observed in purine metabolism, galactose metabolism, and the Krebs cycle; in JIA, in mannose and fructose metabolism, tyrosine, and galactose; in tics, in pyrimidine, fructose and mannose metabolism, and pterin; and in endometriosis, in pyrimidine, amino sugar, and galactose metabolism. Based on ROC analysis, we proposed biological markers for each disease that best distinguish one group from the others. Using the random forest algorithm, we developed a classification model which, based on the complete metabolic profile and metadata (age, sex, body mass index, urine pH and conductivity), enabled differentiation between the five groups with an accuracy of 0.90. Among the most important predictive variables, in addition to metadata, were 2-furoate, trehalose, mannose, and cinnamate. The results indicate that the urinary metabolome enables identification of novel biological markers and the development of screening models for early detection of selected diseases.
|
