How is class code X is identified in stranded data?

[srisarya]

Hi, First, thanks for this utility, it's really useful for my project!

I assembled my transcriptome with Stringtie, and this resulted in predicted transcripts that GFFcompare identifies as relating to a transcript annotated in the UCSC mm39 reference GTF. However, GFFcompare identifies it with class code X, therefore on the opposite strand to the reference transcript it 'belongs' to. How exactly did GFFcompare know this, since it does not take in my BAM (from STAR, by the way, stranded PE short read RNA-seq)?

I can easily enough toggle conservativeness settings with Stringtie, to confirm that this X class code transcript is or isn't an artefact, but I would be running blind, so if there's a way to know what GFFcompare has done this could be helpful. Apologies if I missed something fundamental!

Perhaps it would be good to have GFFcompare add information to the GTF (I'm not sure how) with an edited strandedness according to the predicted class-code?

Thanks!