Bismark: Problem with coverage2cytosine

[AnnaS]

Hi all,
I want to add the strand information on the methylation calls from bismark v0.13 by using the coverage2cytosine script.

Here is an example:
The original methylation calls file (no strand information)
chr1 3000573 3000573 84.6153846153846 22 4
chr1 3000725 3000725 50.7462686567164 68 66
chr1 3000900 3000900 26.3157894736842 5 14
chr1 3001346 3001346 50 2 2
chr1 3001394 3001394 100 2 0
chr1 3002176 3002176 100 3 0
chr1 3002177 3002177 90 9 1

After transforming the data with the coverage2cytosine script I have the information for the strand but the rest of the data are not the same:
chr1 3000574 + 0 0 CG CGG
chr1 3000575 - 0 0 CG CGC
chr1 3000726 + 0 0 CG CGT
chr1 3000727 - 0 0 CG CGC
chr1 3000901 + 0 0 CG CGG
chr1 3000902 - 0 0 CG CGC
chr1 3001346 + 2 2 CG CGT
chr1 3001347 - 0 0 CG CGT
chr1 3001394 + 2 0 CG CGG
chr1 3001395 - 0 0 CG CGC
chr1 3001631 + 0 0 CG CGT
chr1 3001632 - 0 0 CG CGA
chr1 3002177 + 9 1 CG CGG

Do you have any idea of what might happen her?

Thanks in advance,
Anna

[fkrueger]

Hi Anna,

To me it looks like the data in the coverage report is a subset of the original data. The coverage report is based on the genomic sequence, so whenever there is a CG in the reference genome you get a line of report. The first value that is matching is at position 3001346 (marked in bold), so my guess is that you have either a coverage input file which has CG and nonCG values mixed, or you have aligned the data with Bismark/Bowtie2 mode and got some positions which had their context changed to CpG due to deletions in the read adjacent to the cytosine site. These positions are subsequently not picked up in the CpG report since the reference genome simply looks different at these positions. Is either of these a possibility?

The original methylation calls file (no strand information)
chr1 3000573 3000573 84.6153846153846 22 4
chr1 3000725 3000725 50.7462686567164 68 66
chr1 3000900 3000900 26.3157894736842 5 14
chr1 3001346 3001346 50 2 2
chr1 3001394 3001394 100 2 0
chr1 3002176 3002176 100 3 0
chr1 3002177 3002177 90 9 1
...

CpG report
chr1 3000574 + 0 0 CG CGG
chr1 3000575 - 0 0 CG CGC
chr1 3000726 + 0 0 CG CGT
chr1 3000727 - 0 0 CG CGC
chr1 3000901 + 0 0 CG CGG
chr1 3000902 - 0 0 CG CGC
chr1 3001346 + 2 2 CG CGT
chr1 3001347 - 0 0 CG CGT
...

[AnnaS]

Thanks for the reply.

The alignment was made with bowtie, so I think the most probable explanation is that the original file contains CpG and non-CpG sites, although the filename is (CpG_Sample_A_1_val_1.fq_bismark_pe.bedGraph).
The original file was extracted long ago, so I don't know if in the meantime there were some changes in the coverage2cytosine script that could affect the result.
For example, it is strange that very often the original file refers to a base X and X+1 and the coverage report to the X+1 and X+2.
In addition, in most cases the symmetric C on the - strand has zero coverage...

Here, some examples:
chr1 3004635 3004635 75 3 1 --> A on this position (+)
chr1 3004636 3004636 76.7295597484277 122 37 --> C (+)
chr1 3004681 3004681 100 8 0
chr1 3004682 3004682 70.253164556962 111 47
chr1 3004691 3004691 90 9 1
chr1 3004692 3004692 91.6030534351145 120 11
chr1 3004779 3004779 90 9 1
chr1 3004780 3004780 90.6976744186046 78 8
chr1 3004821 3004821 75 3 1
chr1 3004822 3004822 65.8536585365854 27 14

And the cytosine report
chr1 3004636 + 122 37 CG CGA
chr1 3004637 - 0 0 CG CGT
chr1 3004682 + 111 47 CG CGT
chr1 3004683 - 0 0 CG CGA
chr1 3004692 + 120 11 CG CGG
chr1 3004693 - 0 0 CG CGG
chr1 3004780 + 78 8 CG CGG
chr1 3004781 - 0 0 CG CGT
chr1 3004822 + 27 14 CG CGG
chr1 3004823 - 0 0 CG CGT

Thanks again,
Anna

[fkrueger]

If the bedGraph file is very old there is indeed a chance that the positions were 0 based and not 1-based as they are in the coverage file (note that the input for coverage2cytosine is the coverage (.cov) file and not the bedGraph file.

You could simply run bismark2bedGraph again on all the CpG* files that were produced by the methylation extractor, and then run cytosine2coverage again on the .cov file. Good luck!

[AnnaS]

This makes sense...

Thanks a lot!