Are there any parameters that need to be tuned for mapping WGBS data to repeat regions?

[QianhuiWan]

I found that the repeat regions were not well aligned using the default parameters when I aligned the WGBS data to the hg38 genome. I am wondering if there are any recommended parameter settings for aligning repeat regions. My code is as follows:

{abismal} -B -v -i  {hg38_p14_abismal} \
-o ./{sample}_pe.bam -t 16 \
-stats ./{sample}_abismal_stat.yaml \
{sample}_val_1.fq.gz {sample}_val_2.fq.gz

Thank you and best regards,
Qianhui

[andrewdavidsmith]

Repeat regions align well in hg38 -- depending on the kind of repeat. There are no tunable parameters related to this in abismal, but the length of the reads and paried-end helps. I've written multiple papers where analysis of retrotransposons has been important, and mapping of the kind done with abismal usually is fine. It's been fine with most bisulfite mapping tools for a long time. Among the >11,000 methylomes shown in MethBase there are many for hg38 that have covered almost all the CpG sites in the hg38 reference (should be at patch 14), which includes Alus, L1s, etc. The mapping for MethBase was done with abismal using default parameters. This would not include centromeric satellites or many ribosomal regions in acrocentric chromosomes -- but those are not present in hg38 anyway afaik. If you are having issues with mapping repeats, I would first suggest to look at whether the genome you started with has masked repeats. Then if you are still having issues, and your reads aren't unusually short, it would likely be focused on some very extreme subfamily of evolutionarily recent repeats that are very similar to each other. In such a case we can try to find out what are the problem areas and see if there's a fix.

[QianhuiWan]

Hi Smith,

Thank you so much for your prompt reply; that’s very helpful! I am having an issue with the alignment of some L1PA3 elements. They align better with Bowtie1 but show fewer reads after switching to Abismal. The images below illustrate this:
(1) with Bowtie1:

(2) with abismal:

I think this is probably because L1PA3 elements are not considered young LINE1s.

[andrewdavidsmith]

There's no concept of "considered." There is only an issue of whether or not the sequences would be ambiguous. Can you send me the genomic loci of these coordinates? I can't read them from the images.

[QianhuiWan]

Hi Smith,

Thank you so much for your prompt reply! The genomic coordinate is chr8:57100000-57210000 in hg38 reference genome.

[andrewdavidsmith]

I'm attaching a browser image that should be good resolution if you need to zoom in. It shows a bunch of somewhat random methylomes generated using abismal. It's possible a bug has been introduced very recently, and I will check on that, but as you can see from the image, abismal -- at least the versions that have been used to generate this data -- can map almost all the CpG sites within that region. The track across the top shows the locations of CpG sits, and in some methylomes there are gaps. But it seems like most gaps are sample-specific, possibly the individual or cell line has deletions. In general all sites through that interval seem mappable using abismal.

You can browse many thousand methylomes generated in the same way using the link I included above. Just don't try to turn on too many at once because it might cause issues for load time in your browser.

chr8_57100000-57210000.pdf

[QianhuiWan]

Hi Smith, Thank you so much for the information; it’s been extremely helpful! I am currently using Abismal version 3.2.3, but I will try switching to a different version to see if it makes a difference.
Best regards,
Qianhui

[andrewdavidsmith]

@QianhuiWan Honestly I'm not hopeful that it would make a difference. The image I showed you had mostly reads mapped with abismal v1.4.3 and earlier. You are using dnmtools v3.2.3 which includes abismal v1.4.3. If you update to dnmtools v3.2.4 it will include abismal v1.4.4, and I don't recall there being a substantial difference in the code that should affect what you are seeing.

There is one though I had. I haven't tested it. But your reference genome might include alternative chromosomes in addition to the primary chromosomes. For example, if you look at my image all the way at the bottom there is a track named "Reference Assembly Fix Patch Sequence Alignments". It shows alignments with sequences named like chr8_KV88.... If you download certain reference assemblies, those sequences will come with them. And those sequences within the reference would ensure mapping ambiguity. It's worth doing grep chr in your reference genome .fa file to see if there are more chromosomes listed than the 1-22,X,Y and M. If there is anything else, it can cause problems for ambiguity.

[QianhuiWan]

@QianhuiWan Honestly I'm not hopeful that it would make a difference. The image I showed you had mostly reads mapped with abismal v1.4.3 and earlier. You are using dnmtools v3.2.3 which includes abismal v1.4.3. If you update to dnmtools v3.2.4 it will include abismal v1.4.4, and I don't recall there being a substantial difference in the code that should affect what you are seeing.

There is one though I had. I haven't tested it. But your reference genome might include alternative chromosomes in addition to the primary chromosomes. For example, if you look at my image all the way at the bottom there is a track named "Reference Assembly Fix Patch Sequence Alignments". It shows alignments with sequences named like chr8_KV88.... If you download certain reference assemblies, those sequences will come with them. And those sequences within the reference would ensure mapping ambiguity. It's worth doing grep chr in your reference genome .fa file to see if there are more chromosomes listed than the 1-22,X,Y and M. If there is anything else, it can cause problems for ambiguity.

Hi Smith,

I have removed the patches in hg38.p14, and the reads are now aligning correctly to chr8:57100000-57210000. It appears that the issue was indeed caused by the patches. Thank you very much for your guidance and support with this matter—I truly appreciate it!

Best regards,
Qianhui