Question: Bismark methylation for non-uniquely mapping BS-seq reads

Question

I am looking for the DNA methylation, specifically inside small RNAs; piRNA. I am mapping BS-seq reads on the genome with bismark. From the user guide of bismark, it maps reads uniquely on the genome, which mean it reports first read with best score and discard rest of the alignment and if a read mapping to multiple location with same score it will discard all such reads.

Typical command for obtaining uniquely mapped reads from bismark

~/Software/Bismark-0.22.3/bismark GenomePATH/Original_Genome/  --bowtie2 -p 10 -1 Methylation_reads/Sample1_R1.fq.gz -2 Methylation_reads/Sample1_R2.fq.gz --bam -o Sample1

Since, I am interested in piRNA only which may have multiple source in the genome (one piRNA coded from different location of genome), I wanted to take all mapped locations in consideration. Can you suggest what changes I have to make in bismark command. After reading bismark user guide I am confused because I could not able to find relevant parameter.

Also, please let me know if this approach is correct, if not what may be best approach for this.

Answer 1

I'm not certain about your piRNA approach, maybe others will be able to help with that question. As for multi-mapped reads, from Bismark's documentation you can find the --ambig-bam flag:

For reads that have multiple alignments a random alignment is written out to a special file ending in .ambiguous.bam. The alignments are in Bowtie2 format and do not any contain Bismark specific entries such as the methylation call etc. These ambiguous BAM files are intended to be used as coverage estimators for variant callers.

It doesn't look like Bismark will return all of the multiple alignments, just a randomly selected one for a given read or read pair (this probably results from the way Bowtie2 handles multiple alignments within Bismark).