Rna-Seq Library Design With Different Sequencing Depth

Question

[clausndh]

Hi Biostar Experts,

For the first time, I'm involved in designing a RNA-Seq experiment for a Illumina HiSeq-2000.
We have 16 samples which we have to divide at only 6 lanes! So is totally clear that we have to multiplex. And we can't use lane 7! or an additional flow cell.

My idea was to multiplex 4 lanes with 3 samples and 2 lanes with 2 samples. 1. Question: Is this a good design? 2. Question: With these design we will get samples with different sample sizes coming from the different sequencing depths. Do you think this a (big) problem for differential expression analyses?

My belief is that the normalizations form DESeq or edgeR are good enough to solve this problem.

Thank's for your help.

Answer 1

[Devon Ryan]

FYI, you have 16 samples, not probes (in case you're German, "Probe" in German is different from "probe" in English).

In an ideal world you'd just multiplex all the samples on each of the lanes (i.e., put all 16 samples on each of the lanes). Should that prove logistically difficult, your alternative proposal would be OK. DESeq/edgeR/etc. won't have much problem with that (I usually see people running into problems when the size factors differ by ~10x). BTW, depending on your goals, half to a third of a HiSeq lane per sample may be overkill (you can probably get away with 2-4 lanes for your samples, depending on the circumstances).

Answer 2

[support]

If you evenly distribute your samples across 16 lanes, you should get around ~70M reads per sample. See this query on Genohub. If you're looking for differential expression the number of reads you're getting is probably overkill. Here are some recommended parameters.

I agree with the last answer that ideally you should barcode each of your libraries, pool them together and load equally in each lane.

-- Genohub