Figure 5

Heatmaps and confusion matrix for the study group ‘Total’. Data were obtained using the study group ‘Total’ (pooled data from the three centers; 20 control group (CG), 26 osteoarthritis (OA), 33 rheumatoid arthritis (RA)) as the training set for the rule generation and re-applying the respective rules to the same dataset. (A) Heatmap of the membership weights applying all rules of the primary rule set (α = 0.95; ‘CG’, 281 rules; ‘OA’, 25 rules; ‘RA’, 108 rules) for the prediction of the clinical status of the different samples; dashed red lines indicate the lower limits of the respective pruned lists of rules subsequently applied in (B). (B) Heatmap of the membership weights applying pruned lists of rules (α = 0.95; ‘CG’, 21 top-ranked rules; ‘OA’, nine top-ranked rules; ‘RA’, 15 top-ranked rules) for optimized prediction of the clinical status of the different samples. (C) Confusion matrix for the rule set displayed in heatmap (B). TP, true positives; TN, true negatives; FP, false positives; FN, false negatives.