Bisulfite Sequencing Alignment: SOAP, BOWTIE2, BWA, LAST, GEM

Question

[Shicheng Guo]

Is there any comparison of the performance between SOAP, BWA, BOTIE2, LAST, GEM in bisulfite-sequencing data alignment?

What does the bolded part mean? Why do we need unique matches?

The failure of RMAPBS to align reads that contained adaptor sequence, even if these were trimmed, reduces its value as an aligner for RRBS protocols, since it loses the information from shorter reads. For our data, the newer version of BSMAP (v1.2) is considerably improved, since it performs over 40 times faster than the older version. The outputs from BSMAP and RMAPBS require further downstream processing in order to obtain unique matches and to generate CpG methylation data.

Answer 1

[Devon Ryan]

Some of those (e.g., bwa and bowtie2) would perform very very poorly on BSseq data, which is why there are wrappers that perform the necessary magic to use them. For BWA and bowtie2, they can both produce similar quality (search for BWA-meth, which uses BWA and where the github page contains a comparison between it and bison, which I wrote and uses bowtie2).

I've not personally tried SOAP or GEM (maybe I used GEM, I can't recall 100%). With LAST, I tried to use it but recall needing an absurd amount of memory for mammalian sized genomes (the nodes on the cluster I was using only had a quarter terabyte each...which was apparently not enough).

Answer 2

[David Langenberger]

Check out this publication: http://bioinformatics.oxfordjournals.org/content/28/13/1698.long

There is a benchmarking, but keep in mind that it was done for the publication of the tool from the paper. And, since it is from 2012, some older tools were used.

[David Langenberger]

And this publication from 2014 also compares bisulfite mappers.

[David Langenberger]

And this master thesis also shows some comparisons.