RNAseq strand-specific vs.non strand-specific

[Gonza]

Hi there 'RNA-sequers',

I am pretty new to this field and I would like to ask (something maybe basic) if anyone could please explain to me why you'd use a strand-specific protocol to build cDNA libraries out of total RNA instead of normal protocol (i.g., non strand specific one). I am struggling to see what are the benefits for using each, and definitely struggling to understand the biology foundation behind. I mean, you won't see more gene expression depending whether you use a strand specific protocol, right???.

My organism (petunia) does not have a reference genome, would a strand-specific protocol be a better fit???

Thanks very much!!.
Gonzalo.

[dpryan]

With a strand-specific method, you benefit from knowing whether the read originated from the + or - strand. This is particularly useful for finding unannotated genes and ncRNAs. On the negative side, I think it ends up costing a bit more. For those of us just looking for annotated differentially expressed genes in well annotated organisms (particularly mouse and human), strand-specific protocols may not be worth it.

If you're sequencing the genome as well and plan to use the RNAseq data to bolster gene-calling, you'd do well to use a strand-specific protocol. If not, and you're looking at differences due to environment/fertilizer/whatever then someone with experience in de novo assembly would need to chime in, since I have zero experience with that.

[kmcarr]

On the negative side, I think it ends up costing a bit more.

For Illumina TruSeq RNA kits the cost difference between stand-speicfic or not is negligible. I don't have information about other kits. I will agree that there may be some applications where the additional strand information doesn't add much to the analysis, but I really can't think of any situation where having strand-specific data would be detrimental.*

(*Unless the strand specific protocol biases transcript presence/abundance in the the final library but thus far I have heard no reports of this happening.)

[chadn737]

I would definitely favor strand specific data anymore over non-strand specific. Without a reference genome, you will need to de novo assembly and some assemblers work best with strand-specific data, such as Trinity.

[Gonza]

Thanks all for your feedback.
I think it may have to do with the fact that a strand-specific protocol works ONLY with coding regions - CDS (which is what I am interested in) so I can rule out other non-coding RNA's (i.e., rule out antisense RNA, regulatory RNA, etc). This will allow me to see gene expression, not antisense regulation, for example. Any thoughts on this?

...thanks...
-G

[chadn737]

Strand specific protocols are not coding sequence only. Rather they have the added advantage that you can distinguish reads anti-sense reads from the coding sequence since you know the strand that it came from.

[RachelJo]

Thanks kmcarr.

[kmcarr]

RachelJo,

Sorry but that's not true. While "Trinity performs best with strand-specific data" (see http://trinityrnaseq.sourceforge.net/#running_trinity) it will assemble non strand-specific reads just fine.