Is it better to use a merged transcriptome or the reference annotation for RSEM gene expression

Question

[danson.wooi]

Just wondering if it is better to use a merged transcriptome of the samples I have or should I use the human reference annotation? I intend to do differential expression so I was planning on using RSEM to calculate counts. I know that the merged transcriptome might result in novel transcripts but I was wondering if it still would be a fair analysis to just use the human reference annotation?

Answer 1

[JC]

Depends if you want to check only known genes/transcripts or you want to consider the presence of new unknown transcripts/isoforms. From the research point of view you want to do both, then compare the differences and see if there are any advantages from one annotation to the other.

Answer 2

[swbarnes2]

For human, the reference annotation should be better than what you could work out experimentally. So if you aren't looking for novel transcripts, just use the annotation that already exists.