What is the basic principle behind rmats detection of individual events?

[happypiggyzjx]

Hello to all of you on the development team, I have been analysing the results and have become interested and questioned the rationale behind the rmats software determining the 5 different variable clipping events. For example, for MXE events, does the software need to detect 6 loci to confirm an MXE event? For all events except MXE events, does it detect 4 loci to confirm the existence of the event?

The above is just my personal speculation based on the principle of variable clipping event, I am still looking forward to your professional answer, I hope the more detailed the better, or can you tell me where the relevant code is (maybe I can learn it!)

[EricKutschera]

In order to detect an event rMATS needs to have evidence for all of the exons and splice junctions shown in the diagram: https://github.com/Xinglab/rmats-turbo/tree/v4.3.0?tab=readme-ov-file#output

This post has some discussion of how rMATS detects events: https://github.com/Xinglab/rmats-turbo/issues/161#issuecomment-974119283

[happypiggyzjx]

In order to detect an event rMATS needs to have evidence for all of the exons and splice junctions shown in the diagram: https://github.com/Xinglab/rmats-turbo/tree/v4.3.0?tab=readme-ov-file#output

This post has some discussion of how rMATS detects events: #161 (comment)

Thank you for your reply, I have previously studied this literature that your link points to, may I ask if I can roughly understand the following statement: 6 pieces of evidence are needed to determine an mxe event, while 4 pieces of evidence are needed to determine a se\ri\a3ss\a5ss event.

As I understand it, 2 exons (4 if you count the black exons on both sides of the picture) are needed to accurately detect an mxe event, whereas the rest of the events only require 1 exon (3 if you count the black exons on both sides of the picture) to be accurately detected. If this is the correct understanding, can it be argued that there is more variation in detecting mxe events? Also, can I assume that mxe events make up a relatively minimal percentage of all events (I tried looking up the literature but couldn't find data to support this)?

[EricKutschera]

From the diagram for an MXE event, there are 4 exons and 4 splice junctions. All of the other event types have 3 exons. An SE event has 3 splice junctions. For A5SS and A3SS events there are only 2 splice junctions and the other red line in the diagram is for reads crossing the boundary between the short and long exon. RI events have 1 splice junction and the two red lines are for reads crossing into the intron

From figure 2 of https://doi.org/10.1038/s41596-023-00944-2 about 30% of the events are MXE

In supplementary table 2 of http://dx.doi.org/10.1073/pnas.1419161111 about 20% of the events are MXE