High non-CG methylation and low mapping efficiency in PBAT library

[yub]

Hi,
I have a PBAT library from human cells that's spiked with unmethylated lambda DNA and then sequenced on Illumina HiSeq with 150bp SE. The bisulfite conversion rate of lambda DNA is 99%, however, the non-CG methylation in the human cell PBAT library is ~50%. Can anyone explain this BS conversion rate discrepancy in the same reaction?

Also, the mapping efficiency of human PBAT library is very low ~3%. Standard Trim_Glore and Bismark (Bowtie2) alignment was used. Thank you for contributing ideas and solutions!!

[dpryan]

The weird non-CpG methylation is highly likely to be related to the low mapping rate. Bismark has a handy --pbat option, which you'll need to use. If you didn't use that, then the results you see are pretty much expected.