124 with UC, which showed only moderate discriminative capacity (top five: AUC = 0.73, F1-score = 0.71; top ten: AUC = 0.69, F1-score = 0.67) (Figure 5E–F), similar to that observed when using all entries of antibody-bound peptides. Among the top contributing antibody-bound peptides here were an autolysin of Lactobacillus salivarius (N-acetylmuramoyl-L-alanine amidase), a plant fungal allergen (Alternaria alternata) and flagellins from Eubacterium and Legionellaceae and the P200 protein of Mycoplasma pneumoniae (all more frequent in CD) and a Bacillus phage peptide and viral peptides including Norwalk virus, EBV and enterovirus B (all more frequent in UC). Apart from the individual antibody-bound peptides among these top selections, these classification analyses confirm that antibody-bound peptides from flagellin-coated bacterial species dominated among the antibody-bound peptides contributing to the classification of patients with CD from healthy individuals and from patients with UC (Table S10, Table S12). Figure 5 | Discriminative capacity of the top five and top ten antibody-bound peptides between patients with CD and healthy controls, patients with UC and healthy controls and patients with CD and UC using recursive feature elimination (RFE). (A, C, E) ROC curves demonstrating the discriminative capacity of the top five and top ten antibody-bound peptides in classifying between patients with CD and HC (A), patients with UC and HC (C) and patients with CD and UC (E) in the test set (20% of the data). (B, D, F) RFE plots demonstrating the association between the number of antibody-bound peptides entering the classification model and Cohen’s Kappa value, representing a balanced metric between positive and negative predictive values. Abbreviations: ABP, antibody-bound peptides; AUC, area under the curve; CD, Crohn’s disease; HC, (age- and sex-matched) healthy controls; UC, ulcerative colitis. Chapter 4
RkJQdWJsaXNoZXIy MjY0ODMw