This technically pertains to the command line featureCounts program in subread, but the subread webpage suggests posting here, and I'm assuming the bioconductor version has the same functionality.
I'm trying to count only fragments that span splice junctions (per gene as the meta feature) in featureCounts using the --splitOnly option. The documentation suggests that this will capture reads with an 'N' in the cigar string, or individual reads that have split alignments.
However, I am using paired-end data, and I'd like to make sure that I capture fragments (also using the -p option) where the individual mates are not split, but the whole fragment aligns to more than one exon. It's possible I missed it, but I couldn't find anything in the documentation to suggest what the program does in these cases. I checked some of the counts against splice junction counts from STAR, and in a couple of cases it did appear that STAR was generating substantially higher count numbers. I realize that there is a separate option to count reads/fragments spanning individual splice junctions, but I'm not looking for that granularity here.
Thanks so much for any guidance!