Skip to main content
Fig. 1 | BMC Medical Genomics

Fig. 1

From: Pan-cancer analysis of differential DNA methylation patterns

Fig. 1

Illustration of the PanDM model. a Distributions for the empirical null hypothesis \(\left (\mathcal {N}_{c0}\right)\) and the alternative hypothesis \(\left (\mathcal {N}_{c1}\right)\) in different cancer types. b In this toy example, there are nine CpG sites which can be grouped into three clusters. The cluster proportion corresponds to the π in our model. The DM tendency in the three cancer types, termed pan-cancer DM patterns, are summarized in the probability matrix Q by PanDM. c The DM status matrix H derived from Q. Each element of H is a binary variable, indicating whether a specific CpG site is DM or non-DM in every cancer type. d Marginal distributions of the observed data. The distribution results from the mixture of \(\mathcal {N}_{c0}\) and \(\mathcal {N}_{c1}\), where the proportion of each component is given by H. e The z-values for each CpG site in every cancer type. The z-value matrix can be further transformed into the observed p-value matrix

Back to article page