Altering rmap files in specific genomic regions

[Lindsey Montefiori]

Hello,

We are working with capture Hi-C data sets where we have designed baits to all human promoters. In most cases, each promoter has at least two MboI fragments that contain a bait due to our design strategy. This results in interactions with distinct ends over promoter regions, leading to some redundancy in our interactions, thus we would like to limit each promoter region to just one "fragment."

As a test, we merged MboI fragments that contain a bait over the promoter and generated a new rmap file that essentially has one long fragment over each promoter. Obviously this results in a larger number of reads mapping to each promoter fragment since now a fragment represents two or three MboI fragments.

Chicago finds ~2/3 of the same interactions with both rmap files (original MboI and altered promoter MboI), but there is a substantial number of interactions that are different between them. Can you provide any suggestions or insights as to how this manipulation may affect how Chicago calls interactions and whether this approach is correct or contains problems?

Thank you very much.

Lindsey

[Jonathan Cairns]

Hi Lindsey,

Seems like a totally reasonable approach to me. If merging two baited fragments into a single bait, then the s_j estimate for that bait should increase roughly 2x. If you are doing this for the vast majority of baits, as it seems you are, then you won't see a change in s_j, rather a change in f(d).

Using plotBaits() on the baits in question may help explain the discrepancies that you describe, I would imagine they are just down to averaging out the variation between the two baited fragments.

Thanks,

Jonathan

[Lindsey Montefiori]

Hi Jonathan,

Thanks for the reply. Is it not the case that it will now be 'easier' to find promoter-promoter interactions since there are essentially 2X as many reads over promoters compared to other ends when we merge promoter MboI fragments? In the regular Chicago run, we find ~5% promoter-promoter interactions, however this goes up to 20-30% in the Chicago runs with the larger promoter regions.

Thanks, Lindsey

[Jonathan Cairns]

Hi Lindsey,

Yes I think it is easier to find promoter-promoter interactions, since if both promoters have merged fragments then the number of reads goes up 4x! Meanwhile, most interactions' read counts go up only 2x, so one would expect the percentage of significant promoter-promoter interactions to go up.

Thanks,

Jonathan