Fig. 2

a Heatmap of methylation β-values for the 6 differentially methylated CpG sites used for classification analysis. The 6 CpG sites shown are required and sufficient for perfect linear discriminant classification of samples with fibrotic progression (FP) and stable disease (SD). The order of samples equals their class membership probability and is identical to the order of samples in b. Additional file 4: Figure S4 shows the results of the hierarchical clustering based on these 6 CpG sites separately for the training and test cohort. b Linear discriminant classification of samples based on 6 differentially methylated CpG sites. Linear discriminant values (LD1) show the predicted class membership of samples with fibrotic progression (FP, left) and stable disease (SD, right). Colored background areas indicate regions of high confidence (dark, class probability ≥ 95%) or lower confidence (bright, class probability > 50%). The linear discriminant classifier was trained using methylation β-values from 6 CpG sites of 22 samples (colored sample labels). Subsequent model validation was performed against 11 independently measured samples (white sample labels).