[Q] Centromere Trans Pileup

[LucasMcNU]

Hi all,
I was wondering if it's possible to reproduce the trans centromeric pileup plots from Figure 1b in this paper with Coolpup. I tried with our data (<5kb resolution, human cells) and it seems that the paucity of contacts at the centromeres from repeat mapping exclusion makes those regions difficult to analyze using pileups (my guess).

I used the centromere/pericentromeric coordinates for hg38 on UCSC. I tried rescaling and plotting the raw pileup, in addition to adding a flanking region. Here is an example of a result:

produced with this code (as an example, i've tried a few different things):

condA=MB
res=5
threads=12

coolA=/projects/b1103/HIC_Macquarrie/Lucas/merged_cools/${condA}/${condA}_${res}kb_merged_filt.cool
expA=/projects/b1103/HIC_Macquarrie/Lucas/merged_cools/${condA}/contacts/${condA}.expected.trans.${res}kb.tsv
out_fileA=/projects/b1103/HIC_Macquarrie/Lucas/merged_cools/${condA}/contacts/

dir_ann=/projects/b1103/HIC_Macquarrie/Lucas/annotation/
dir_res=/projects/b1103/HIC_Macquarrie/Lucas/loops/

coolpup.py --nproc $threads $coolA ${dir_ann}mb_centromere_chr2.bed --features_format bedpe --expected $expA rescale --rescale_flank 5000000 --rescale_size 99 --trans --outname ${out_fileA}${condA}_5kb_trans_cent_pileup.clpy
plotpup.py --input_pups ${out_fileA}${condA}_5kb_trans_cent_pileup.clpy --plot_ticks --output ${out_fileA}${condA}_5kb_trans_cent_pileup.png

[Phlya]

This should be totally possible, but the figure in question shows, from what I understand, whole chromosome-chromosome pileups. So with pileups it's a little tricky, since the positions of centromeres vary across chromosomes... You have to create rescaled pileups separately for all arm-arm interactions, and then put them together manually into one matrix. I think this is what was done in the paper, just not using coolpup.py, but @agalitsyna maybe can correct me here.

Probably you can use saddle plots from cooltools instead, might be easier... If you create an artificial track with the coordinate relative to centromere-telomere distance, and use it instead of the eigenvector there.

[LucasMcNU]

so if I understand right, the centromeric pileups were generated by piling up entire arms separately and then averaging individual pileups for each arm in a single matrix? They used cooltools for that particular plot. I'd love to hear more from @agalitsyna if possible. I've seen some examples of telomeric trans pileups for coolpuppy but has anyone done something similar for centromeres? Thank you again?