I have made 'reference' transcriptomes from several different isolates. I have tried mapping my RNAseq reads with the tuxedo suite to these transcriptomes in order to conduct DE. Some transcripts that we experimentally know to be present are getting FPKM values of 0. It seems this is either due to multi-mapping of reads (E.g. many similar isoforms) or relatively low abundance (if there are other possibilities please let me know).
I am wondering what you all think about repeating the above but breaking my reference transcriptomes up into smaller subsets (to reduce multi-mapping). I suspect this would greatly increase FPKM values (giving more of a max value then a real value) but it might be possible to include a reference transcript (or several) in each subset for normalization.
Is this crazy? I just can't work with my original data, which is telling my transcripts I know to be present are not.
Any other ideas?
I am wondering what you all think about repeating the above but breaking my reference transcriptomes up into smaller subsets (to reduce multi-mapping). I suspect this would greatly increase FPKM values (giving more of a max value then a real value) but it might be possible to include a reference transcript (or several) in each subset for normalization.
Is this crazy? I just can't work with my original data, which is telling my transcripts I know to be present are not.
Any other ideas?