Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Questions about TAD calling #27

Closed
Rseq opened this issue Aug 14, 2020 · 4 comments
Closed

Questions about TAD calling #27

Rseq opened this issue Aug 14, 2020 · 4 comments

Comments

@Rseq
Copy link

Rseq commented Aug 14, 2020

Hi,

First, I would like to thank you for developing FANC tools and make it available.
It is helpful and it made simpler to analyse Hi-C data.
However, I'm once again asking for your help.

I intend to perform TAD calling and I would like to know what is the difference between fanc boundaries and tadtool.
It was not clear for me in this statement if fanc boundaries is a 1-TAD caller algorithm itself or 2-calls tadtool

We are therefore currently not bundling a TAD calling tool with FAN-C, and refer the user to one of the many available tools for TAD calling that offer a wide range of features. https://vaquerizaslab.github.io/fanc/fanc-executable/fanc-analyse-hic/domains.html#a-note-on-tad-calling

In case it is option 1- a TAD caller algorithm itself, which one has performed better on your hands?

Thank you for your time

@kaukrise
Copy link
Collaborator

Hi, thank you for your question.

fanc boundaries is not a TAD caller. It calls insulating boundaries - individual bins with a strong insulating potential, i.e. comparatively few contacts spanning them. TADs are domains which are delineated by those boundaries. As stated in the documentation quote you linked to above, FAN-C does not come bundled with a TAD caller, and it also does not call TADtool.

TADtool uses the same algorithm to call insulating boundaries, and then has an additional step where it determines if any region between two boundaries has sufficient signal enrichment above background to be a TAD.

TAD calling is very sensitive to parameter choices (insulation score window size, boundary strength cutoff, matrix resolution, etc), and also depends on the types of TADs you are interested in (large domains, nested domains). Therefore we have written TADtool, which allows you to find robust parameter ranges visually. However, our work has recently focused more on TAD boundaries, or the insulation score around genomic elements of interest, than TADs themselves.

When you want to work with TADs directly, I would advise you to use multiple of the many available TAD callers, preferably with different algorithms, and get a robust consensus set of domains. Otherwise, depending on your question, it may be preferable to work with the insulation score directly, rather than the domains.

I hope this is helpful!

@Rseq
Copy link
Author

Rseq commented Aug 14, 2020

Thank you for your prompt reply.
It was quite helpful! I misunderstood it, that is why I did this confusion..
However, please correct me:
If

FAN-C provides multiple “scores” that are designed to find the boundaries between domains. https://vaquerizaslab.github.io/fanc/fanc-executable/fanc-analyse-hic/domains.html#hi-c-domain-analysis

Then, once one knows the boundaries, would not be TADs what is in-between?

I believe that for my aim, to check if there is a difference in TAD borders among different conditions, the insulating boundaries are even better.

Thank you again!

@kaukrise
Copy link
Collaborator

kaukrise commented Aug 14, 2020

Then, once one knows the boundaries, would not be TADs what is in-between?

Consider these examples:

              /\
       /\    /##\
      /##\/\/####\/\
TADs: --- -- ---- --
      
       /\
      /##\________/\
TADs: ---         --

In the first one, the TADs are indeed all regions between boundaries. That would also mean that every single base in a genome is part of a TAD. However, there are often large genomic regions with weak or no discernible structure between boundaries. I suppose it is a matter of definition if you want to call them TADs or not, but this is why TADtool requires a minimum level of signal between boundaries before a TAD is called.

@Rseq
Copy link
Author

Rseq commented Aug 14, 2020

I suppose it is a matter of definition if you want to call them TADs or not

I agree with this concept if one considers the in-between in the second example as unique-extended boundaries

TADtool requires a minimum level of signal between boundaries before a TAD is called

However, my perception is closer to this as well

Thank you again for your time and clarification!
All the best

@Rseq Rseq closed this as completed Aug 14, 2020
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants