Fig. 6

Dealing with biases and artefacts. a A screenshot from the Bismark ‘bam2nuc’ module output, showing base and dinucleotide content in a Heat dataset against the genomic expected value. The C base indicates the extent of degradation-caused bias (negative correlation), the G-base and its derivative dinucleotides as well as the A/T bases and dinucleotides show the extent and direction of amplification bias—G(C)- or AT-biasing. b Comparison of two methylation quantification strategies to overcome or decrease the effects of GC- or AT-biasing protocols. Counting a total number of methylation calls within a probe (region), irrespective of their position and depth, is compared to calculating methylation values of individual CGs and averaging those for the whole probe. None of those approaches applies initial coverage depth filtering, which is shown in Additional file 2: Figure S9. c–e Non-CG methylation in the major satellite after removing conversion noise by c setting a 10 or 20% 5mC cut-off threshold value, d after a bioinformatic filter was applied to remove every read with three or more methylated CH cytosines, and e after subtracting the background values from an unmethylated genome control (TKO mESC). Positive y-axis values indicate the top strand and negative the bottom strand