Motif specific, polyploid DMR analysis?

[jkh00]

Hi,

as implied in the title, I'm wondering if a --motif parameter can be implemented into the dmr analysis?

I wanted to know if a specific region between my two samples is differentially methylated in a context specific manner (CG, CHG and CHH etc). I performed pileup using the following command and ran dmr pair on the bedmethyl.gz. The result is however, a pooled of all the contexts. My current workaround is splitting the bedmethyls into context specific and then run dmr pair on those splitted bedmethyls.

My questions are:

1. Is this a proper way to detect dmr for each motif?
2. and could this be parameterised in modkit dmr?

My code to run modkit pileup

modkit pileup --motif CHH 0 --motif CHG 0 --motif CG 0 --combine-mods --log-filepath pileup.log --threads 48 --ref TAIR10_genomic.fna input_aligned.bam output.bed

Thanks in advance!

[jkh00]

Another unrelated question:

3. if I have a polyploid sample (tetraploid for example), and I would like to know the DMRs between these haplotypes. I have a phased genome, should I do mapping on the phased genome, perform pileup, split accordingly to the haplotypes and perform dmr multi on these bedmethyls?

[ArtRand]

Hello @jkh00,

Unfortunately, Modkit doesn't have an "out of the box" solution for tetraploid genomes. It does have the tools to do what you need, but you will have to do the steps manually.

First what I would do is align your reads to the unphased genome but retain the haplotype tags. Then partition the BAM by haplotype so you'd have something like this:

mappings_unphased.bam
mappings_hp1.bam
mappings_hp2.bam
mappings_hp3.bam
mappings_hp4.bam

This way all of the reads are on the same coordinate system so you can look for differentially methylated Cytosines.

Then using Modkit v0.6.0 run pileup at each motif like so:

for hp in "1" "2" "3" "4"; do
  for motif in "CG 0" "CHG 0" "CHH 0"; do
    motif_fn="${motif// /_}"
    pileup=pileup_${motif_fn}.bedmethyl.gz
    modkit pileup mappings_hp${hp}.bam $pileup \
      --modified-bases 5mC 5hmC \
      --ref $ref \
      --motif ${motif} \
      --bgzf
    tabix $pileup
  done
done

Finally, do modkit dmr pair with for each pair (HP1 vs HP2, HP1 vs HP3, ..., etc.). You can run modkit dmr multi if you have regions of interest that you want to compare. All dmr multi does is run a loop over all of the pairwise comparisons.

Sorry there isn't a slicker solution, but this will get you tables of the diferentially methylated cytosines, separated by context between each of the haplotypes. You can then look at the distribtion of scores and find the ones in the 95th percentile or rank my MAP-based p-value, etc. Hope this helps, let me know if you have any more questions.

[jkh00]

Thanks a lot for the workaround solution @ArtRand, I will try it out :))
Aside from the polyploid issue, do you think a motif specific parameter could be implement into modkit dmr pair/ multi in the future?

[ArtRand]

Hello @jkh00,

Let me look into it - but it's probably doable since you're essentially subsetting the sites in the bedmethyl. The DMR module is being given a "re-think" so I can evaluate this feature as well.