Identification of candidate biomarkers for inflammatory bowel disease using non-targeted serum metabolomics

Objective. To identify candidate biomarkers for inflammatory bowel disease (IBD) using non-targeted serum metabolomics in patients with ulcerative colitis (UC). Patients and methods. This pilot study included 9 male patients with active UC and 11 healthy male controls. Serum metabolomic analysis was carried out by gas chromatography-mass spectrometry (GC-MS). Compounds were identified using the NIST08 mass spectral library. Classification of samples and search for candidate biomarkers were performed using a support vector machine (SVM), projections to latent structures-discriminant analysis (PLS-DA), and a naive Bayes classifier (Naïve Bayes). When choosing a classifier, we were guided by the areas under the ROC curve. Results. Metabolomic analysis revealed 85 compounds, of which 79 were annotated. The normalized serum levels of 14 metabolites (2-hydroxybutyric acid, caprylic acid, erythronic acid, creatinine, β-glycerophosphate, α-glycerophosphate, 2-keto-D-gluconic acid, pentadecanoic acid, trans-palmitoleic acid, palmitic acid, margaric acid, palmitoleic acid, squalene, α-tocopherol) differed significantly between the groups. Serum levels of all these compounds (except α-tocopherol) were elevated in UC patients compared to healthy controls, which was accompanied by an increase in the concentration scatter. The largest area under the ROC curve corresponded to the Naïve Bayes classifier (AUC = 0.931; excellent model quality). Conclusion. The study identified 14 metabolites that can be used as candidate biomarkers for IBD after proper verification, both for diagnosis and for assessment of the efficacy of therapy. The identification of disease-associated metabolites will facilitate the development of novel biotherapeutics. Key words: inflammatory bowel disease, ulcerative colitis, metabolome, biomarkers, gas chromatography-mass spectrometry.