Figure 3

Derivation of a minimal, multiply-redundant, methylation signature for medulloblastoma subgroup identification. Using a three-class (a) and two-class (b) subgrouping model, a minimal multiply-redundant methylation signature was identified by selecting a subset of signature candidates with the highest ranking in a classifier fusion model (support vector machine (SVM), artificial neural network (ANN), decision tree (DT) and Bayesian network (BN). (c) Heatmap of the most discriminatory signature candidates (n = 17) recapitulates the four known medulloblastoma subgroups. Chromosomal locations (hg19 assembly) and identifiers of the 17 CpG loci are reported in the adjacent table.