search

dieterich-lab / circtools

circtools: a modular, python-based framework for circRNA-related tools that unifies several functionalities in a single, command line driven software.
http://circ.tools
GNU General Public License v3.0
25 stars 20 forks source link

Different results for circtest from circtools or from R #69

Open idotu opened 5 years ago

idotu commented 5 years ago

When I run circtest with the same data from R or from circtools I get different results, how is this possible? Relevant info:

  1. circtools: circtools circtest -d mydir/ -p 0.0001 -s 1 -f 0.99 -C 1 -r 3 -g 1,1,1,2,2,2 -l CondA,CondB -c 4,5,6,7,8,9 -o out_dir/

  2. In R: CircRNACount_filtered <- Circ.filter(circ = CircRNACount, linear = LinearCount, Nreplicates = 3, filter.sample = 1, filter.count = 1, percentage = 0.0001)

test=Circ.test(CircRNACount_filtered,LinearCount_filtered,CircCoordinates_filtered,group=c(rep(1,3),rep(2,3)),alpha=0.99)

Thanks,

ivan

tjakobi commented 5 years ago

Additional information via mail:

I've re-checked it and it's something odd, the difference I see is just in one circ that with circtools shows as significant with sig_p 0.0026412332380743 and with circtest in R it is not (sig_p = 0.428199220581546). Somehow this seems wrong because it only has 1 read in one replicate, 0 everywhere else.

tjakobi commented 5 years ago

Dear @idotu,

would it be possible to transfer the CircRNACount and LinearCount files from DCC in order to reproduce the error?

How many columns / samples do you have in your CircRNACount files?

Cheers, Tobias