Hello,

I am very new to R and have spent a lot of time looking at the DESeq2 guides and package overviews etc, but have not been able to normally run any scripts due to the experimental design of our project being too complicated - especially for a newby in bioinformatics.

We have RNAseq count data from an experiment where we have 9 samples - 3 groups of 3 samples - A(1,2,3), B(1,2,3), C(1,2,3) - all should be in the same condition but they are samples from different sequence backgrounds - bacterial samples from different clonal complexes.

Each of the samples have 3 biological replicates = 27 lanes in the RNASeq library - and these have been sequenced twice, a second set on a different set of lanes giving us 2 technical replicate of each biological replicate = 54 lanes.

I am struggling to find a way to collapse the technical and biological replicates. Ideally I would like to have a step where I would be able to: a. normalise the count data and do some statistical comparisons of the technical replicates (n=2) for each biological replicate to show any differences (as DE would not be statistically possible) to see how much technical error we have introduced to the data set. b. normalise the count data and run DE & statistical comparison between the biological replicates (n=3; A(1(1,2,3))...) for each sample (n=9) to see how much our biological replicates vary (these would be pairwise & clustered (?)) c. run differential expression and see DE genes within each group A, B, C. d. run differential expression and see DE genes across groups / cross comparison A,B,C.

I am able to run simple pairwise comparisons by HAND for each biological/technical replicate as I haven't figured out a way to automate the scripts/inputs. I am not sure if pairwise comparisons are the right way to do it as overall it will not give us the whole picture overview we would like to see (and is really tedious).

Any and all help would be appreciated! (and please don't link me to the package index).

Thanks for the help!

Technical replicates should be combined using collapseReplicates, see the vignette. This leaves biological replicates only.

Then, if you are interested in the effect of A-C, and want to control for 1-3 (I don't know what the name of these variables is for you) you would use a design such as ~clone + condition. For looking at how much the biological replicates vary, I suggest taking a look at the vignette and workflow examples, e.g. PCA plots, distance matrices, and so on.