Choosing matrix normalization method for .cool files

[rikrdo89]

For cool files that have values for different types of matrix normalization (KR, VC, VC_SQRT, etc), is there a way to specify which normalization values to use? I tried /path/to/file.cool::/resolutions/5000::/bins/KR but that didn't workl...

resolutions
     ├── 5000
     │   ├── bins
     │   │   ├── KR (617665,) float64
     │   │   ├── VC (617665,) float64
     │   │   ├── VC_SQRT (617665,) float64
     │   │   ├── chrom (617665,) int32
     │   │   ├── end (617665,) int32
     │   │   ├── start (617665,) int32
     │   │   └── weight (617665,) float64
     │   ├── chroms
     │   │   ├── length (24,) int32
     │   │   └── name (24,) |S64
     │   ├── indexes
     │   │   ├── bin1_offset (617666,) int64
     │   │   └── chrom_offset (25,) int64
     │   └── pixels
     │       ├── bin1_id (495409844,) int64
     │       ├── bin2_id (495409844,) int64
     │       └── count (495409844,) int32

[lldelisle]

Hi,
This is a good idea of enhancement (but needs to modify the package we depends on: HiCMatrix). For the moment, we do not support it so you need to use hicConvertFormat to extract the values with the normalization you want. You need to pay attention to the argument --correction_division depending where your matrix comes from. If you need help with this, I can help you.

[rikrdo89]

Thank you @lldelisle. I am back working with HiC matrices, and I would appreciate if you could share an example on how to extract the KR normalized values for plotting the matrix with pygenometracks.

I started with a .hic file from HiC-Pro; and then converted it to .cool using hicConvertFormat -m "$myHic" --resolutions 5000 --inputFormat hic --outputFormat cool -o "cool/${myHic/.hic/.cool}". When looking at the metadata using hiInfo, I see that the cool file contains KR ,VC, and VC_SQRT normalization values.

What would be the next step to properly get a cool file with the KR values?

[lldelisle]

Hi,
Sorry for the late reply.
I know it is possible. Could you share a toy dataset similar to yours?