Publicly Available Pipeline For Analysis Of Whole Genome Bisulfite Sequencing Data

Question

[skm770]

Hi all,

I am trying to find any publicly available pipeline for whole genome bisulfite sequencing. Although there are a couple of them that I have found but either they are not properly documented (as is the case with Bsmooth (http://rafalab.jhsph.edu/bsmooth/) or they don't seem to do what I want like in the case of SAAP-RRBS (which is mainly used for RRBS and uses a different aligner (BSMAP) than the one I want ( Bismark). I was hoping if anyone knows about a pipeline which uses BISMARK for whole genome bisulfite sequencing.

regards Saad

[Matt Shirley]

What more do you want than is already provided by Bismark? The user guide provides all information necessary for running the program to provide methylation summarization, which can be converted to bedGraph using the provided script, and then visualized in your genome browser of choice. Are you interested in identifying differentially methylated regions?

[skm770]

I have different kinds of data sets belonging to illumina/ABI-solid and I was hoping there could be a pipeline which starts from fastq files till giving DMR

[skm770]

How about something that includes qc manipulation as well??

[Devon Ryan]

There are multiple levels of QC that could be done. Firstly, you can just add fastQC to your pipeline. Secondly, bismark can now create output for an M-bias plot, which looks informative. Bismark also writes its alignment metrics to a text file, so you can ensure that replicates are similar. Those are the standard QC metrics, though one could come up with more (e.g., if there was a phiX spike-in, then you could align to that to calculate the bisulfite conversion percentage).

Answer 1

[Devon Ryan]

Why don't you just align with bismark (I'd suggest my own aligner, which is faster, but I see you also have SOLiD data), use bismark_methylation_extractor to make a bedGraph file for your CpG sites, and then use that as the input to bsseq (which can be thought of as the analysis component to bsmooth)? Making a pipeline to get everything ready for analysis in R would be quite simple.

[Chris Whelan]

Yes, I've used this strategy too, and it worked well.

Answer 2

[Charles Warden]

I think methylKit is pretty good, and I'm pretty sure it works with Bismark (or at least it shouldn't be too hard to modify the percentage methylation files):

http://code.google.com/p/methylkit/

If you were willing to focus on a select subset of CpG islands, you could also use COHCAP (which is specifically designed to work with Bismark):

http://sourceforge.net/projects/cohcap/