RNA Seq differing Read Number

[Greenhorn]

Hi,

First off all, I want to thank you for this nice forum in general. I already learned a lot!

I´ve got a question concerning differing readnumbers between biological replicates. We want to analyse differential expression in two different tissues from maize.

First we started with one replicate only, to test the whole procedure and to have a first glance on potential results.
After a evaluation of this first dataset, we decided to go on and sequenced additional replicates afterwards.

so, now we have one replicate with around 70 million reads and another replicate with around 30 million reads. I did some checks and we are quite sure that 30 million reads are sufficient to do the anylsis. So would you be concerned about the different read numbers? Or would you try to sample the first one down?

As far as I understood, the RPKM value would take this into account. But I am not sure if it is also ok for the EdgeR analysis, since it uses total counts, right?

Sorry if this question is redundant or stupid; but I am really a Greenhorn

[blancha]

The first step performed by edgeR is normalization of the samples relative to the sequencing depth (i.e. total number of reads). This "is always present, and doesn’t require any user intervention." (edgeR manual)

On an unrelated note, you should be wary of any batch effect when analyzing samples that were not processed together. The batch effect is definitely less of an issue in next-generation sequencing than for microarrays, but it may still be present.

[Greenhorn]

Thank you very much blancha! That is what i hoped to hear .

We are aware of potential batch effects and did a correlation analysis to check this. The correlation between the replicates in tissue 1 is 0,9 which is not great; but ok, since the praparation of this tissue is not trivial. The replicates of Tissue 2 show a correlation of 0,95 (both based in RPKM).