FDR = 1 for known splicing event

[ebonner303]

Salutations,

I am analyzing a cohort of publicly available RNA sequencing data (taken from multiple sources). I study a specific mutation, which leads to several well established splicing changes. When I run RMATs on the large cohort, the FDR=1 for many of the well established splicing changes. I have tried to remove samples with low quality or read count, but I am still encountering this error. when I look at the IJC and SJC for one of the A3SS events i see:

IJC_SAMPLE_2

250,230,191,205,89,152,119,140,29,25,34,30,56,19,54,2,3,4,0,0,0,0,1,0,0,0,0

SJC_SAMPLE_2

65,94,72,64,13,53,25,37,3,8,16,6,9,4,0,0,0,0,0,0,0,0,0,0,0,0,0

Is the issue all the zeros? I would really like to do my downstream analysis on events with an FDR<0.05. I dont know if this is just an issue of using publicly available data or if there is some way to overcome this?

[EricKutschera]

It looks like 8 of the samples have no reads for the event. Both IJC and SJC for those samples are 0. There is a branch with code to ignore those samples with zero reads: #318

You could check the read_outcomes_by_bam.txt in the --tmp directory to see if the reads for those samples are being filtered out for some reason