No significant events when 70 control samples were used

[Akshaya Vs]

Hey Eric,

Thank you so much for the tool. I was wondering if there the significant event would be affected if the number of comparing samples are bit skewed.  I initially ran rMATs on samples 5 samples with a particular variant vs 70 controls (samples without the variant), and this resulted in no significant event whereas when I down sampled the controls to 5 samples, I got around 3k FDR < 0.05 SE events.

I was wondering if such a big skew in the number of controls does not make sense in the first place.

[kutsc...@gmail.com]

Having different numbers of samples in the two groups should be fine. One possible reason for the 70 controls having no significant events is that one or more of those samples have zero reads for events that would otherwise be significant: https://github.com/Xinglab/rmats-turbo/issues/318

Another possibility is that there is a lot of variation among the 70 controls which leads to the events not being significant. If you take some of the significant events from the 5 control output and then find those events in the 70 control output you can check if there are samples with zero reads (IJC + SJC = 0) or if the IncLevel values vary a lot

Eric