Need help with methylation bias

Question

[Antonio R. Franco]

II need your advice

I am analyzing by WBGS (whole genome bisulfite sequencing) the genome of my plant under two conditions with 3 biological replicates each. I did run a Illumina sequencing of paired reads, and mapped the reads with both, bismark and bwameth looking for methylation in the three plant contexts, that is CpG, CHH and CHG

I am sort of confused because the methylation bias I got using any of these two mappers. Pictures shown below show that R1 reads have a nice mbias easy to supersede, whereas R2 show a increased or decreased bias along all the reads in all my samples

So I have several questions

1. Why this "crazy" mbias is systematically present in all my R2 reads

2. Is it a nice idea or not to do paired mapping when analyzing methylation?

3. Any idea on how to treat these three biological replicates ?. Should I average the data ?

Answer 1

[Devon Ryan]

1. My guess is that the base quality of read 2 follows a similar curve, such that the early bases have very poor quality. Aside from that I don't have a good guess.
2. It's usually better to use PE reads.
3. This depends on your goals, but normally you would continue treating them as replicates when you get to the DMR calling stage (assuming that's what you want to do).