Performance of protein levels for multivariable classification of MC. Machine learning classification, performed by penalized logistic regression in 30 nested cross-validations, and their corresponding model variable importance scores for (A); MC active, MC-HR, and all MC vs. controls, and (B); MC active, MC-HR, and all MC vs. UC active. The left section shows the averaged receiver operating characteristic (ROC) curve of nests with 95% CI (mean area under the curve [AUC] and 95% CI shown in legend parentheses) for predictions of MC. The boxplots show the model importance scores of the top-10 proteins selected by at least 80% of model fits in the nested cross-validations for (A); MC active and all MC vs. controls and (B); MC active and all MC vs. UC active. The left and right borders of the boxes indicate the first and third quartiles, the lines in the middle represent the median, and the whiskers extending to the most extreme points within 1.5 times the interquartile range (IQR).MC, microscopic colitis; HR, histological remission; UC, ulcerative colitis.