I am doing a project in RNA-seq data with some datasets that are paired-end data. After alignment with TopHat2, I have around 1-5% of reads whose mate is mapped to another chromosome. In order to call Cufflinks for annotation of transcripts, would you recommend me to remove those reads mapped to another chromosome (or call again TopHat2 with the option --no-discordant)? Is it a key step?
Thanks in advanced.