Are exon skipping events that are missing from single samples indicative of no skipping evidence?

[Ryan Corbett]

We have analyzed alternative splicing events in a large pediatric cancer data set using rMATs turbo with the --statoff argument to only calculate IncLevel1 in each sample without replication. 

We have merged results across all samples and see that, for many exon skipping events, IncLevel1 values of 1 and SJC_sample values of 0 are reported indicating no reads supporting exon skipping. However, in many cases, exon skipping events are not reported in samples even when we can see them in sashimi plots. For example, the TSC exon 11 skipping event (chr16:2060669-2060813, upstream exon chr16:2058746-2058873, downstream exon chr16:2061870-2062008) is reported in rMATs output from the first sample in the sashimi plot below, but it is not found in any of the other samples even though there are high read counts at the junctions supporting inclusion. This leads to our question--will rMATs always report exon skipping events even when the IncLevel1 value is 1, or there certain situations when events with no evidence of skipping are not included? 

Thanks in advance!

[kutsc...@gmail.com]

rMATS will only detect an event if there is evidence for all exons and junctions in the event: https://github.com/Xinglab/rmats-turbo/issues/265#issuecomment-1927162591

If a sample doesn't have any reads to support the skipping junction and the --gtf also doesn't define the skipping junction then rMATS will not detect that event when run with only that sample. Since you ran with --statoff, all detected events are reported even if the sample doesn't have supporting reads, but the event still needs to be detected

In the example it looks like the skipping junction is not in any annotated transcript. In that case rMATS needs to see a read supporting the skipping junction in order to detect the event

If you run multiple samples together in the same --task post (or --task both) then rMATS can use the reads from all samples together to detect events: https://github.com/Xinglab/rmats-turbo/tree/v4.3.0?tab=readme-ov-file#running-prep-and-post-separately

Eric

[Ryan Corbett]

This is very helpful and clears up our confusion. Thank you for your prompt response!