study design for RNA-seq

[equus]

I am considering using Illumina for RNA-seq to measure gene expression in a prokaryotic genome of ~ 6Mb. Cost is an issue, but I am trying to make sure I take into account biological and technical replicates, as I have seen many comments about the importance of replicates, etc.

My experiment will yield 8 prokaryotic samples with 3 biological replicates for each sample. I am thinking that to save money I can prepare the libraries myself using the TruSeq kit and order extra barcodes so each biological replicate would have its own barcode, for a total of 24 barcodes that I could sequence in one lane. Then I would perform my technical repeat in a second lane of sequencing on the same libraries. My question is whether or not that would that be a sound technical repeat? Can I use the same libraries between technical repeats? Also, any recommendations on how many reads per sample? I just saw a paper saying 500 million or more for eukaryotic, but what about prokaryotic? Also, is there any value to paired-end reads versus single-end reads for prokaryotic RNA-seq?

I am new to NGS, and I must apologize for how basic some of my questions may be--any advice is most welcome!

[kmcarr]

Equus,

I'll share my point of view with respect to your questions:

1. Biological replicates are essential for analysis of differential expression and I think your plan for 3 biological replicates is fine. Technical replicates are of much lesser value and I never recommend them to researchers. You are doubling your sequencing costs for only marginal (if any) gain in value.

2. Illumina just announced their TruSeq kits will be upgraded to 24 barcodes so you are covered there.

3. I'm not sure where you saw a recommendation for 500 million reads per sample but that is crazy talk, for either prokaryotic or eukaryotic. I typically will recommend an initial target of ~20 million reads per sample and that is usually for complex eukaryotes. For a small prokaryote 10 million reads per sample will probably be a safe place to start (it's always easy to add more reads once the libraries are made).

4. If you already have an annotated reference genome and will therefore only need to align your reads to it then absolutely single-read will be fine. For a prokaryotic genome they don't even need to be that long (very little repetitive sequence). Single reads of 50bp (or even 36 if the budget is that tight) will be good.
If you do not have a reference genome and need to assemble a transcriptome de novo from your reads then you should consider longer reads. Paired end would help in this case but for a prokaryotic transcriptome (no splicing, isoforms, etc.) you could get away without pairs.

5. Are you sure making the libraries yourself will save a significant amount of money? You can't simply compare the list price of the TruSeq kit to the price for sample prep your sequencing provider quotes. There are consumables required and not included in the kit to take into account. As well as other equipment/supplies you might not have on hand. Not to mention the expertise the sequencing lab has already built up in preparing these libraries. Determine what the real cost to do it yourself is and ask if any savings are worth the time and effort you will have to expend doing it.

[GenoMax]

Equus,

Do not forget to account for a per sample cost to analyze the data, if you can't (or are not planning to) do the analysis yourself.

[equus]

Dear Kmcarr,

Thank you for the detailed and informative responses to my questions--I will definitely take all of your comments into account. Indeed, you pointed out some important things to consider. I will carefully weigh all of the factors when deciding whether or not to make the libraries myself--I am not sure about how much time it may take to optimize and all the extra materials it takes, etc.

Thank you!

[equus]

Equus,

Do not forget to account for a per sample cost to analyze the data, if you can't (or are not planning to) do the analysis yourself.

That is a good point.

Thank you