Understanding "TotalEvents" in "TotalEventsJC" and "TotalEventsJCEC"
X L
Hi All,
I am trying to understand the “TotalEvents” values in the “TotalEventsJC” and “TotalEventsJCEC” columns of the rMATS summary file (attached). I consistently observe that the “SE” event type has the highest count in both “TotalEventsJC” and “TotalEventsJCEC” compared to other event types.
I assume that “TotalEvents” represents the number of events annotated in the GTF file that are also detected in my RNA-seq data. Interestingly, “SE” events also have the highest counts in the “SignificantEvents” columns (both JC and JCEC). While this may not be surprising due to the high overall counts of SE events, I am considering using the ratio of “SignificantEvents” to “TotalEvents” for each event type (e.g., SignificantEventsJC of SE / TotalEventsJC of SE) as a more meaningful indicator of which event types are most affected by treatment.
For example, based on the attached summary file, rather than concluding that “SE” is the most influenced splicing event after knockout of factor X, it might actually be the “A3ss” event type when considering these ratios. From what I’ve seen in the literature, many rMATS users report SE as the most affected event type, but this may not hold true if these proportional ratios are taken into account.
I would appreciate any insights or input from anyone familiar with this.
Best,
Xiao
Thomas Danhorn
It is not surprising that SE has the most events, because it is the only
category that does not rely on annotated alternative splice forms. E.g., for
a RI event to be tested, your GTF has to have a transcript with that
intron retained (besides the "regular" one). Same for A5SS, A3SS, and
MEX. Potential SE events, on the other hand, are automatically contructed
by simply skipping every "middle" exon (i.e. anything that can be skipped
by splicing).
I don't know if taking the ratio of actual to "total"/potential events is
a great idea, though (although it is tempting), because you would be
comparing ratios with denominators that were derived from completely
different premises -- in the SE case, that would be everything that is
possible, whereas in the other cases it would be restricted to annotated
events, which presumably either have been observed or at least predicted
based on some evidence. Unfortunately there is no good way I can think of
to compare the releative prevalence of one event type to others in this
scenario.
Best,
Thomas
> You received this message because you are subscribed to the Google Groups "rMATS User Group" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to rmats-user-gro...@googlegroups.com.
> To view this discussion visit https://groups.google.com/d/msgid/rmats-user-group/221796e3-e84c-4f0f-a2be-23a01f3cebden%40googlegroups.com.
>
X L
Hi Tom,
Thank you very much for your input. Instead of focusing on which event types are most affected in treatment versus control, perhaps we can concentrate on the changes in the usage of individual splice sites or splice junctions (donor-acceptor pairs), since all event types identified by rMATS are manifestations of differential splice site or junction usage.
Best,
Xiao
Thomas Danhorn
control and treatment. I have done something along these lines before
(https://journals.plos.org/plosgenetics/article?id=10.1371/journal.pgen.1004932).
There are several different variables you can look at (besides PPT and
exon length, there are enhancers, branch sites, etc.), you have to see
what gives interesting results.