search

4dn-dcic / hic2cool

Lightweight converter between hic and cool contact matrices.
MIT License
63 stars 7 forks source link

Balancing weight inversion, or not #26

Closed Phlya closed 5 years ago

Phlya commented 5 years ago

There have been issues about it already, but still. As far as I understand, hic2cool>=0.5 should not invert KR etc normalization weights, and they should be divisive. But I have just converted a bunch of .hic files from Rao et al 2014 paper, and the weights are clearly multiplicative: image Note that weights are below 1, but in the info it is 'generated-by': 'hic2cool-0.5.1',. Am I misunderstanding something?

Phlya commented 5 years ago

Apologies, if I go further away form the telomere the weights look centered around 1 and not very small numbers, and looking at the heatmaps, the data is balanced when I use these weights (with the new cooler that special-cases these names). All is working as expected.