insulation score on large domain all negative values

[jiangshan529]

.
Hi, I am using cooltools insulation to calculate insulation score at 5kb resolution, the parameters I used are cooltools insulation a.mcool::resolutions/5000 -p 8 --bigwig -o out --ignore-diags 3 125000. but in a large compartment domain, the insulation score is all negative, which really looks weird to me. what can I do to at least calculate the scores at those regions be around 0. Thanks.

[Phlya]

Clearly in this sample this whole domain is highly depleted of interactions... Possibly it's a heterozygous deletion or some huge accessibility bias against getting contacts from there (and/or balancing didn't quite work to fix the coverage). Insulation score is normalized to the whole genome (or chromosome? Not sure atm), and so super depleted region will look like that...

[jiangshan529]

hi, Phyla. Thanks for the quick reply! It's from two different WT cell types. the first figure here is a zoomed out region for the previous example at 25kb resolution without normalization. and second figure here is a more conserved region in the same chromosome. I think that region is a heterochromatin in hct. The third figure is KR normalized, should not be a problem for normalization. Do you think it's fair to say the blue circled region is a cell-type specific (sub)TAD? And by the way, does cooltools insulation by default use ICE normalized matrix based on my command? Thanks!

[Phlya]

With HCT116 cells I would not be surprised if indeed it is a deletion, they don't exactly have a completely intact genome. Looking at the first map, it looks like the "corner dot" in HCT has this weird shape characteristic of hemizygous deletion, but hard to be sure just from this image.

However heterochromatin is indeed an alternative explanation, as it seems like these are micro-C maps? MNase digestion can be very biased by accessibility and indeed coverage of heterochromatin is lower.

You can check published Hi-C data for this region and see if it also shows the same pattern or not.

Otherwise, yeah, the circled region looks like a cell type specific boundary.

[jiangshan529]

Hi, Phyla, for this region I do not worry about the incomplete genome, since there are reads can be mapped. and also this is around centromere. the second plot is from HCT insitu hic. Yes, I also have the experience before that insulation score calculated from insitu hic looks better than microc. Do you think it's worth to fix the issue for microc insulation score calculation?

[jiangshan529]

this is insulation score called from insitu hic. I can see in the large compartment domain it's still all negatives even though I am just using 25 window for sliding.