mRNAseq, single or paired?

[pfranchini]

Hi,

We are running a RNA-seq experiment to evaluate differential gene expression in different biological groups using Illumina. We already have some transcriptomic data for this species generated with a 454 sequencer.
Is it necessary to sequence paired-end data or, because we already have some good reference, single-end is the best strategy to use. I know that paired-end reads are better for de novo assembly, but I was wondering if they could have some advantages to evaluate gene expression profiles apart from the assembly step. Moreover, even if we have some transcriptomic data, the fact to have paired-end data and create a mixed assembly (with reference and de novo) could help to catch more ESTs. In conclusion, our doubt is if it is worth to spend more money to generate paired-end reads considering the benefits we could have for our experiment.

Thanks

[shurjo]

I would definitely go with PE as that is the general direction RNA-Seq data analysis software seems to be taking. In the first year or so after RNA-Seq was introduced, PE data was generally just treated as single end unless your lab had some talented bioinformatics folks available. These days a lot more off-the-shelf software actually uses the PE information (which is indeed a powerful layer of the data), so I use PE runs for all my libraries now.

[kwatts59]

I got very good results for differential expression on a plant genome for samples exposed to various treatments using single end 50bp reads.
If DE is all you need to do, I think single end would be fine.

[chadn737]

It really depends on what you want to do with the data. Are you just looking for differential expression? If so, SE will likely be enough, if you have other goals, then PE will be better.