I've gone over a couple of previous questions discussing a similar subject and they seem to all imply that it simply is not feasible to do with unstranded RNA-Seq data.
We know that for RNA-Seq, RNA is reverse transcribed into cDNA and that antisense RNA is complementary to its mRNA on the sense strand. If we were interested in detecting antisense RNA for a given gene, can't we just look for a complementary sequence of the gene (e.g. first 20 sequences)?
I understand this could be a problem since we don't know whether the matching sequence from the RNA-Seq reads is actually coming from the antisense strand, but how likely is it to detect, for example, a sequence of length 20, on the sense strand by chance?
Are there any additional things that should be taken into consideration?
Thank you.
De novo discovery of antisense transcripts using unstranded sequencing data can be challenging. One trivial approach is to construct a list of annotated antisense transcripts whose splice sites do not overlap with the corresponding sense transcripts that code for proteins. You can use reads that are mapped to these junctions as a proxy to approximate antisense expression. You can perhaps expand such approach with splice site discovery to identify unannotated antisense. YMMV.