metawaffle to PrognoTF

Requirements

• scipy
• numpy

optional

• TADbit to convert from TADbit BAM format to required
• matplotlib to plot and check the result of tests

Generate sub-matrices at a given resolution

We will need sub-matrices from which we will extract sub-matrices.

To run it we need:

• Genomic interactions in four column format (see bellow)
• one or two lists of coordinates

The genomic matrices are in the form:

    388761 388761 364 0.461182349185
    388761 388762 66 0.166867406676
    388761 388763 142 0.554955487814
    388761 388764 4 0.0216862127515

- First column: starting bin
- Second column: end bin
- Third column: raw count
- Fourth column: normalized count

• resolution

• tmp directory to store the pre-processed files.

• output directory to store the results

• biases (in my case is a pickle file, as the output from TADbit using oneD biases), so if you want to use another one you should modify it.

And we generate:

• matrices per chromosome: folder where the matrices per chromosome are located. The format is the following: Creation of such matrices with bam2count.py.

Notice that the chromosome 1 goes from bin 0 to (length of chromosome 1) / resolution, and chromosome 2 goes from the (length of chromosome 1) / resolution + 1, until (length of chromosome 1) / resolution + 1 + (length of chromosome 2) / resolution. Example: chromosome 1 (length 10), chromosome 2 (length 5)

bins chromosome 1: from 0 -10.
bins chromosome 2: from 11-16.

Now these files have the name: chr{number}_bam_5kb.tsv

• windows span: length of the windows to add around the center of the peak (bp)
• max dist: maximum distance between centres of the peaks.

Generate list of pairs of coordinates

Finally, this code is written to analyze the following intervals of distance between the peaks:

   windows = ((255000  , 1000000),
              (1000000 , 2500000),
              (2500000 , 5000000),
              (5000000 , 10000000),
              (10000000, 15000000),
              (15000000, 20000000),
              (20000000, 30000000))