Help with interpretation of M-bias plot

[xinyun-stacy-li]

Thanks for creating and actively maintaining this amazing tool! I'm a first-time user of Bismark. I'm running this pipeline starting from FastQC, Trim Galore, Bismark aligner, and Bismark methylation extractor. The libraries I'm working on is NEB EM-seq paired-end sequenced samples. I trimmed 10bp from 5' and 3' end for both read 1 and 2 according to your recommendation here (https://felixkrueger.github.io/Bismark/bismark/library_types/#em-seq-neb). The M-bias plots generated from one of the samples look very weird, especially for read 2 where the CpG methylation starts from roughly 40% and wiggles for the first half of the read length. I'm also confused about the reason why CpG total calls in read 1 starts to gradually drop even in the middle of the read length. I wonder if you could help me interpret the plots and suggest potential fixations. Thank you so much!