Unstranded STAR Alignment with XS tags - reverse stranded assignment

[Riley-Grindle]

Hi @gpertea and StringTie Team,

I have been using StringTie to refine annotations of various naive transcriptomes and came across a data set using unstranded RNAseq. The standedness prediction algorithm correctly assigned the reads as unstranded so no --fr and --rf flag was assigned in my pipeline. Additionally, I have set up the STAR aignment to use --outSAMstrandField to intronMotif. After running StrinTie (with -j 4 and -a 15) it appears we are still getting regions with sufficient evidence of reverse stranded splices that are directly overlapping with forward junction coverage but the forward structure are the only annotations being called (see below, forward strand 1st, reverse strand 2nd):

When manually checking this region in the BAM file used for the run it is apparent that the XS tags are present and show both +/- for those junctions.

Any help or guidance on this behavior would be greatly appreciated!

Best,
RIley Grindle