Is this an assembly error in the rice genome? Because a transcript is a copy of one DNA strand (which may then be spliced to remove introns) -- to me this makes having parts of the CDS on both strands impossible. Or are you saying there are two transcript strands, one from each DNA strand, which then get somehow spliced together? Or am I misunderstanding something? Either way, I would be surprised it any of the common tools out there could deal with something like this ... Short of "fixing" the assembly and annotation, the choices I see are • excluding any transcripts with strand

?

, as you have done, or • mapping to the transcriptome (instead of the genome) -- rnaseq can do this with the

--pseudo_aligner

option. The latter gives you all genes, but limited QC (e.g. it is not clear if a read failing to map to the transcriptome is intronic, intergenic, or something else completely [junk, contamination, etc.]), so it's a trade-off. Since the pseudo alignment is very fast, you could potentially do both.