RNA-seq differential analysis with N batches and a technical replicate in each batch

Question

[jonathanli6022]

Hello,

I am looking to perform differential analysis for an experiment with a relatively unusual design.

We have RNA-seq for 60 disease and control samples and would like to conduct differential analysis between the conditions. These samples were generated in 8 separate iPSC differentiation batches, and it would not be surprising to have significant batch effects between batches. Our idea was to include a technical replicate in each of the 8 batches which would give us a sense of the variance among batches.

I've already tried including including batch as a blocking factor in DESEQ/EDGER, but this approach does not take advantage of the replicate in each batch. So my questions are

1. Can DESEQ/EDGER accommodate such an experimental design?
2. If not, are there others that might help?

Thanks!

[Arup Ghosh]

There is a similar thread in bioconductor forum. https://support.bioconductor.org/p/73399/

[Gordon Smyth]

The Bioconductor thread doesn't seem similar to me.

[ting.w.sf]

How did you solve it? I have similar situation with multiple samples sequenced in different times but with two samples as technical replicates for each time.

Answer 1

[Gordon Smyth]

You have included a technical replicate in each batch, but a technical replicate of what? A replicate of the same sample in every batch?

Anyway, including batch as a blocking factor seems the obvious and best approach. You can easily assess variance between batches by testing for differential expression between batches or simply by an MDS plot.

If all the batches include both disease and controls samples, then there is little or nothing to be gained from technical replicates.