Single End vs Paired End

[Wiseone]

I would love some input on this one. I have collected samples from an exposure and my goal is to perform RNA Seq to investigate differential gene expression. As there is no reference genome for my species (Chironomid dilutus in case anyone is wondering) I will have to do a de novo assembly of the transcriptome. My plan is to use all reads from each sample (one sample per lane, 3 lanes) to assemble my transcriptome, and then perfrom expression analysis for each sample. What I need to decide upon is whether to go with Single End or Paired-End sequencing. I realize that the de novo assembly will be "easier" with paired-end reads, but I will lose depth of coverage with paired ends compared to single ends. So, is the de novo assembly that much more reliable with paired-end reads that it outweighs the loss of coverage?

I am rather new to this area so my question may be naive. I continue to read what I can but I find that forums such as this are much better resources than reading the primary literature.

[NextGenSeq]

You will gain coverage not lose it with a PE run. You get twice the amount of data (coverage). It's also twice as expensive and takes twice as long though.

[Wiseone]

Thanks for the reply. If I were to choose single end runs (75bp) is de novo synthesis of the transcriptome effective? I have 454 on the titanium system as an option as well. Unfortunately I cannot afford both and have to make a decision. If de novo synthesis of single end reads is effective I would choose it over 454 based on the amount of data generated.

[james hadfield]

It is not twice as much to run PE, per bp it is significantly cheaper the SE.
I would choose Illumina (but that is what we run) and think your big choice is 36 or 72bp runs? If this is a one off and you want as much data as possible for assembly then my guess is long means better. If the transcript abundance is more important then more but shorter reads would be the way to go.

[Lee Sam]

I vote for paired end as well, though you might want to choose a relatively small insert size. Just be aware that the size selection step in sample prep will keep you from detecting mRNAs below the selection size. I think Illumina has what they call a DSN normalization protocol for total RNA capture though - I remember that they presented at a local user group meeting that they captured a lot more snoRNAs as a result of using that sample prep.

[yuan.li]

I vote for paired end as well, though you might want to choose a relatively small insert size. Just be aware that the size selection step in sample prep will keep you from detecting mRNAs below the selection size. I think Illumina has what they call a DSN normalization protocol for total RNA capture though - I remember that they presented at a local user group meeting that they captured a lot more snoRNAs as a result of using that sample prep.

Does anyone know how well this DSN normalization from Illumina work? Why is snoRNA captured by DSN but not the regular protocol? If size selection keep you from detecting mRNAs below the selection size, how can DSN get around and detect smaller size mRNA? I thought DSN is just to eliminate majority of the ribosomal RNA...

[malachig]

I would also go with paired-end runs. The paired reads will have a higher mapping efficiency, its easier to identify amplification bias, and the insert size itself is extra information with many possible uses. A further practical reason to chose paired-end is that a lot of methods, algorithms, tools, etc. are being developed with paired-end reads in mind...

[Lee Sam]

Does anyone know how well this DSN normalization from Illumina work? Why is snoRNA captured by DSN but not the regular protocol? If size selection keep you from detecting mRNAs below the selection size, how can DSN get around and detect smaller size mRNA? I thought DSN is just to eliminate majority of the ribosomal RNA...

Here's a page with the link for the DSN normalization protocol. I can't comment further because I honestly don't remember the specifics.

http://www.illumina.com/applications...ncing/rna.ilmn