Figure 1.

Schematic of the theoretical principle underlying signal detection applied to methylation analysis. Knowledge of the probability distribution for the methylation signal, here given as the Hellinger Divergence (H) of methylation levels, permits the identification of “true” differentially methylated positions (DMPs) [21,22]. Critical values are used to discriminate between a biological methylation signal and the molecular thermal noise generated by biochemical processes and conforming to laws of statistical physics [25,26,27,40]. Next, signal detection is designed to identify an optimal cutoff value, (here denoted as $H_{33}^{D_T}$) to discriminate the methylation signal induced by the treatment from the natural background variation [5,33]. Empirical comparisons allow the placement of Fisher’s exact test for the discrimination of DMPs.