Fig. 3

Example pycoMeth workflow for differential methylation analysis. A Methylome segmentation using a sBayesian changepoint detection model. Segmentation can be computed on a read-group (e.g., haplotype) level. Emission likelihood in the HMM models methylation call uncertainties as well as an optional methylation rate prior. B Differential methylation testing allows for a number of test choices. The LLR difference hypothesis compares methylation call LLRs within a segment between two samples directly. Selecting the count dependency hypothesis or the \(\beta\)-score difference hypothesis (default) both result in binarization of methylation calls based on a defined LLR threshold. The count dependency hypothesis leads to a test on contingency tables, testing dependency between methylation count and read group, whereas the \(\beta\)-score difference hypothesis results in a test comparing, for each segment, the read methylation rates between read groups. Regardless of test hypothesis, p-values are then subjected to multiple testing correction. C The reporting module generates an overview HTML report, as well as individual interval reports