I am trying to do batch correction on the junction read counts, and I am trying to extract the categorised junction supporting reads for each alternative event. Hence trying to understand how rMATS works a little more.

Why does the individual counts supporting IJCs i.e. for SE upstream_to_target and downstream_to_target not add up to the total IJC value?

i.e.

I know reads can span multiple junctions, so I thought maybe these reads would increment both categories but would only tick the total count by 1? However, looking at IGV for particular examples this doesn't always seem to be the case.

I presume therefore that individual junction counts are separate to the computation of the IJC and some filtering / process goes on?

Some help understanding this would be greatly appreciated!

Thanks