Are there any tools able to exclude specific bases (bias) from methylation calling ?

Question

[hxlei613]

After getting BS-seq alignment files, methylation bias plots can show the methylation bias such as the start of reads in PBAT-Seq experiments. Usually these bias are removed in the following analysis.

I know bismark can remove bias using --ignore. What if I don't use bismark ? It seems that BSeQC can do that too ( I just have a glance at its paper.) I wonder if there are other tools able to do that too.

Thank you very much for sharing ideas.

Answer 1

[Devon Ryan]

PileOMeth can do that as well.

[hxlei613]

I am hesitate to use it because of -D <int> option. Does it mean I need to get exact depth of every base and select the maximum ? The default 2000 may not be suitable for different alignment files.

[Devon Ryan]

This is equivalent to the same option in samtools mpileup and turns out to be a convenient way of excluding huge stacks of PCR duplicates. If you're doing targeted bisulfite sequencing, then raise it to a larger value, otherwise it's fine as is. For RRBS and WGBS, you'd want to ignore sites with a fraction of that coverage anyway.

Answer 2

[igor]

If you know there is a bias, you can trim the reads (FASTQs) before alignment. This way, you don't have to worry if the methylation-specific tools you end up using support trimming.

[hxlei613]

I did think about it. But trimming the reads may result in too short reads, which could waste a number of them. On the other hand, the bias is due to gap repair, trimming may affect mapping efficiency. So I only trim adapter before alignment.When it comes to methylation calling, exclude bias part.