Fig. 6

DNA methylation modeling to predict gene expression. A visual depiction of our simplified model showing the effect of gene body CG methylation and an increasing complexity of interaction terms on gene expression. a A scatterplot comparing Z-transformed gene body CG methylation values with log(gene expression) values. The black line shows the linear term, green line includes both the linear and quiadratic term, and the blue line includes linear, quadratic, and cubic terms. b Interaction plot depicting the interaction between gene CG methylation and exonlength, up-stream CHH methylation, gene body CHH methylation, and gene body CG methylation on gene expression. Summed terms across these four terms are considered ranging from −1.6 (dark purple) to 1.6 (yellow). Points represent actual genes CG gene body methylation, gene expression, and their color represents their interaction sum on the same scale as the model colors. c The second order interaction term of squared gene body CG methylation by exon length is added to the model depicted in b. As exon length increases (goes from red to blue) gene body CG methylation is found to have a more positive effect on gene expression. Points represent genes, and colors represent the exon length of these genes on the same scale as the model colors. d The independent effect of exon length on gene expression is added to the model depicted in c. The shape of the lines does not change, however predicted gene expression is altered (the lines move up or down on the y-axis) depending on the predicted effects of exon length on gene expression