hypermethylated CGs

[JohnUrban]

Hi,

Thanks for the great and ultra-fast tool.

I see that the GFF provides stretches of hypo-methylated regions, but does not distinguish stretches of hyper-methylated regions from regions that did not have enough coverage to call.
For example, a GFF line looks like:

tig00000000     .       epigenetically_modified_region  4009    6343    .       .       .       type=hypomethylated_interval;avg_score=-2.5190;nCpG=97

It would be great if the GFF or separate GFFs included lines about hyper-methylated regions and regions that were ignored -- e.g.:

tig00000000     .       epigenetically_modified_region  1    3008    .       .       .       type=ignored_interval;avg_score=X;nCpG=Y
tig00000000     .       epigenetically_modified_region  3009    4008    .       .       .       type=hypermethylated_interval;avg_score=X;nCpG=Y

I know this type of thing was brought up at least once already in issue #6 . However, I just think it is important enough to bring up again. I suppose I can use complemented regions of the genome to identify 'potential' hyper-methylated stretches, but would need away to remove the ignored stretches. Do you have any recommendations on how to filter the complemented regions to remove CpGs that ignored and to keep only those that have evidence of hyper-methylation? I am looking for an automated/command-line way rather than manually looking at the wig in a browser, for example.

Do you think the wig files could be converted into a BEDtools-friendly format for that type of analysis...?

best,

John

[hacone]

Hi John,

Thanks for the valuable comments.

wig files will be helpful in this case. In wig files, only the positions with coverage at least 3 (1.5 in the representation in file) are reported. To extract the hypermethylated regions, you can just find the stretches of 1 in class.wig file. I think you can write a script to handle the wig files and output hypermethylated regions. I'm not sure there is a general wiggle-to-bed conversion tool available but it's easy too in this case.

Including the hypermethylated regions in gff file seems good idea, and I have a plan to thoroughly improve input/output of the program, so thanks again for the suggestion.