Differential loop detection does not agree/match with read counts between samples.

[narzouni]

Hello,

I am currently using Mustache for loop detection for our MicroC data. Thank you so much for developing Mustache. It is a great tool.

I have decided to use Mustache's Differential loop detection which was introduced to version 1.2.0.

• Do you have any description of the methods for differential loop detection? I looked at the main paper, but I don't think differential loop detection is mentioned in the paper.

• I am asking because I used Mustache's differential loop detection for two samples. Then I extracted the number of reads for each loop (the loops from differential detection) from each sample, but the number of reads does not show or agree statistically with the differential loop detection.

I am sure I may be missing something, that is why I am asking if you can provide more information about your methods for differential detection.

Some of the reads between samples were even contradictory in the sense that the loops which showed as stronger in one sample ( according to differential loop detection) have shown lower number of reads in comparing two samples.

As an example: Loop1 has less read count (24 reads) for sample1, and loop1 has more read count (48 reads) for sample2. However, differential loop detection showed that Loop1 is stronger in sample1 (with FDR 8.82E-07) which is supposedly statistically significant.

I am running differential loop as follows:

     mustache/diff_mustache.py -p 16 -f1 Sample1.mcool -f2 Sample2.mcool -r 4000 -st 0.7 -pt 0.25 -pt2 0.25 -o Sample1_vs_Sample2_r4000_st0.7_pt0.25

Your help will be greatly appreciated.