Merging files with different normalizations and resolutions into a single .hic/.mcool file

[Nuturetree]

Hi
Author
How merged the .hic/.cool files with different resolutions and different normalization
Wariting your reply
Nututures

[robomics]

Hi!

We do not provide such functionality because it would not be trivial to prevent users from creating broken files by providing incompatible inputs to hictk.

Can you elaborate a bit on what your use-case for such functionality would be?
Is it something that cannot be achieved with a combination of hictk load/merge zoomify and balance?

[Nuturetree]

Thank you for your quick response. Let me explain the actual problem I need to solve in detail. Currently,

I have generated 5kb (A_5kb.hic), 10kb (A_10kb.hic), and 20kb (A_20kb.hic) hic files using hictk. Additionally, I have normalized these hic files, resulting in files such as A_5kb_ice.hic, A_10kb_ice.hic, A_20kb_ice.hic, A_5kb_scale.hic, A_10kb_scale.hic, and A_20kb_scale.hic.

May I ask, how can I use hictk's merge functionality to combine these 9 files into a single hic file?

[robomics]

Thanks for providing more details.

I think what you are trying to do can be achieved as follows:

• Generate the base resolution with hictk load
• Create additional resolutions by iterative coarsening with hictk zoomify
• Balance the multi-resolution .hic file generated by the previous step with hictk balance
• Profit!

In practice:

hictk load interactions.pairs.gz -f 4dn A_5kb.hic --bin-size 5000  #  This you already have
hictk zoomify A_5kb.hic A_multires.hic --resolutions 5000 10000 20000  # You can also not specify resolutions and let hictk generate some default ones for you (see docs for `--nice-steps` and` --pow2-steps`)
hictk balance ice A_multires.hic --force  # --force is needed because you have already balanced the base resolution, and so you need to tell hictk to overwrite existing weights

hictk merge is only meant to be used to merge the raw counts from .cool and .hic files that are using the same reference genome.

[Nuturetree]

Can hictk generate a .hic file that contains ICE, scale, and raw statues？

[robomics]

Yes it can!

You can balance the same file (regardless of the format) with different algorithms (and even different variants of the same algorithm).

In your specific case, simply run hictk balance scale A_multires.hic after balancing with ICE as shown in my previous message.
There are various way to speed up loading/balance/zoomify using e.g. multiple CPU cores, more ram and/or fater IO for tempoary files.
Make sure to check out the docs for more details (e.g. the docs for hictk balance are available here).