Questions about RNASeq

[daanum]

Hi,

I have a question regarding the RNASeq.
I am working with Human cell lines. For RNA Seq :
i) Should I consider Paired end sequencing or Single end?

ii) I am planning to get the sequencing done on Illumina for 6 samples in duplicates. Is having replicates for each samples a must? In my case if I consider duplicates then I will have 12 samples all together.

iii) Will it good to pool the 12 samples across 4 lanes in Illumina sequencer? My aim to ask this question is that how much depth are we looking at in case of RNA Seq?

Thank you for your kind attention.

[SylvainL]

Hi daanum,

i) it depends if you want to do only Differential Expression. If the answer is yes, then single-end of 50bp is sufficient. If you want to discover new splicing variants, paired-end is better...

ii) Are you speaking about biological or technical replicates? For RNAseq, biological replicates are much more than recommended. Best is to have at least 3 from my own experience... technical replicates are not of a great use.

iii) For human, and for normal DE, 30 millions reads may be enough in most of the cases. It also depends on which tissue/cells you are working. Is there any over-represented RNA in this tissue/cell? Are you interested in rare mRNA?

In general, all RNAseq experiments are a balance between different variables: sequencing depth, paired-end vs single-end, length of the reads, number of replicates... all of this has a cost and depending on your budget, you may have to make some choices. The cheapest (and still correct way to do RNAseq) is (for DE), single-end of 50bp, 3 biological replicates, 5 samples per lanes (I always count around 150 millions per lane, even though you can expect 250 millions, but this also depends on the platform. The one I am working with, has sometimes quite bad yield...)

[dpryan]

1) It depends on the question you want to ask. If you just want to look at differential expression then don't bother with paired-end reads. If you need to look at splicing events (or allele-specific expression) then paired-end reads might be useful.
2) Describe what you mean by "sample", I suspect that it's not what the rest of us mean. In general, for each biological group of samples that your sequencing you require an absolute minimum of 3 samples. Ideally you would use more (I've typically used 6 or so). If you're asking about technical replicates then don't bother with them unless you have a real need.
3) This depends completely on the biological goal. For differential expression you can run 12 samples on a single lane. For other types of experiments you'll need more.

BTW, you might start a new thread for things like this in the future.

[fanli]

Do you guys find 30M reads to be really necessary for DE? I haven't had much trouble going as low as 10M as long as sufficient replicates.

[dpryan]

The number of replicates is definitely more important than the depth. 10M would probably work fine in most circumstances where there are enough replicates.

[SylvainL]

Yes, I agree. It all depends on your samples (whether you have an abundant mRNA or not, and if you want to look at rare mRNA). As I said, design of RNAseq will also greatly be dependent of the cost: adding samples means additional cost in library prep while, at least in the platform I am mostly working with, increasing the sequencing depth has almost no cost...

[InDel]

1, The type of reads depends on your research goal. For gene expression RNA-seq, single end might be enough. For splicing, gene editing, fusion gene detection, pair end would be better.
2, More biological replicate is always better. No amount of depth is going to provide confidence of the quantification that replicates will give you.
3, The required depth depends on the research goal. For gene expression RNA-seq, 10 M is a minimum. For splicing, gene editing, the more the better, people may use 60M or more. The number of samples per lane also depends on the platform you use.

Another factor you need to consider is poly-A selection vs rRNA depletion for lib preparation.