Cufflinks GA and Directionality

[zorph]

Hello,
I was trying to compare RNA-seq samples run on the solid platform and the GA platform.
I aligned my GA data on using Tophat.
My Solid data was aligned using bioscope.

Then both samples were analyzed using Cufflinks and Cuffcompare.

My question is, on my .loci file, i have strandness information about my GA read that align to non-annotated areas. This should not be possible because my GA library was paired-end, hence no directionality.

My Solid data on the other hand seems to be lacking a lot of strandness information. Edit: my Solid reads were adapted specifically to have directionality.

Can anyone tell me what is going on? Thanks in advance.

[kmcarr]

Zorph,

Cufflinks will impute strand (directionality) to the gene models it predicts based on the canonical directionality if splice junctions. If there are a sufficient number of junction crossing reads in a predicted gene model, cufflinks will assign a strand to that model. It does not require stranded reads to do this since it is the directionality of junction site which is important.

As to why cufflinks is not predicting orientation for the models from your SOLiD data I can't say for sure. Perhaps there at not enough junction reads among your SOLiD data for cufflinks to predict orientation of the underlying gene model.

[GKM]

Are you sure it is not predicting directionality for all the transcripts, i.e. you didn't look at the first X of them and got that impression? It will not give you directionality for transcripts that don't have introns, so if you had a lot of those in the beginning of your file, this could explain it.