map methylome data with GKNO mosaik program

Question

[zhiguangli88]

Hi,

I found the program mosaik in GKNO tool sets (Lee W-P, Stromberg MP, Ward A, Stewart C, Garrison EP, et al. (2014) MOSAIK: A Hash-Based Algorithm for Accurate Next-Generation Sequencing Short- Read Mapping. PLoS ONE 9(3): e90581. doi:10.1371/journal.pone.0090581) is a powerful mapping method. For pair-end reads, it utilizes the knowledge of approximate fragment length in library construction to rescue the ambiguously/multiple anchored read if its mate is uniquely aligned.

I am wondering whether this progam is able to process methylome data generated by MethylC-seq, and whether the above strategy is deployed in processing such data. My datasets were sequenced on HiSeq2000 with PE100 (pair-end 100bp).

Do anyone has any experience with this problem?

Thanks,

Zhiguang Li

Answer 1

[dariober]

My I guess is no, it can't handle BS-Seq data.

If you want to align BS-Seq reads, from directional protocol, with an aligner not designed for BS-Seq data you could follow these steps (but you would probably reinvent the wheel...):

• Convert all the C in reference genome to T. Append this converted genome to the one where all the G are converted to A. I.e. each original chromosome is present twice: as C to T converted and G to A converted. Make chromosome names unique (e.g. by appending to the original name _CT or _GA). Index this reference with your favourite aligner.
• Convert all the C in your reads mate 1 to T. Convert all the G in mate 2 to A
• Align converted reads to the converted reference.
• Parse chromosome names in output SAM/BAM to have the original chromosome names (e.g. by removing _CT or _GA)

(Hope I got it right...)

[brentp]

looks right. at the end, you also need to recover the original reads without the in silico conversions and report those instead of the converted.

[dariober]

Sure... Thanks for pointing it out. (And thanks for bwameth!)

[brentp]

sure, glad someone is using it.

[zhiguangli88]

Dear Dariober,

Thanks very much for your explanation. But looks like it is pretty complicated. Do you know any public available program that can handle pair-end methylome data?

Thanks again.

Zhiguang Li

[dariober]

bismark, which uses bowtie or bowtie2, it's probably the most popular (and for good reason!). Recently I got pretty good results with bwameth.py, which is backed by bwa mem (see also this Bwa-Meth: Align And Tabulate Bs-Seq Reads). Definitively, unless you have a good reason to use a specific aligner, don't reinvent the wheel!

PS: Your comment should have been a "comment" not an answer.

Answer 2

[alistairnward]

As mentioned above, Mosaik is not ideal for your task. I am not familiar with processing methylome data, but you would definitely want to employ a tool that has been developed specifically for handling this type of data. Mosaik was not developed with this in mind. We would be happy to host tools to help parcel together a pipeline, making it a more straightforward task to handle this problem, but the current gkno toolset does not currently contain tools or pipelines for methylome processing.

[zhiguangli88]

Thanks. I will try Bismark to see what I can get.

Answer 3

[Rohit]

Why not try Segemehl which has modified options for Methylation data too. Also for downstream analysis, they released Metilene a few months back.

http://www.bioinf.uni-leipzig.de/Software/metilene/