Generality of the quantitative models. Random Forest expression models were constructed using methylation and expression data from one of the individuals or cell lines, indicated by the different columns. The methylation level of a gene is defined as the average level over its upstream, transcribed and downstream regions. These models were used to predict the expression levels of genes in another individual/cell line, based on their measured DNA methylation levels of them in it. For each of these model training/testing combinations, the prediction accuracy values of the genes in different expression classes, and their overall average, are shown in different bar groups. Within each bar group, the accuracy values based on the four DNA methylation measures are shown.