I have a data set of six sinalge samples with no replica. I know, that DESeq can work with these, as I have done it already. I would like to know, what to do in this case, when working with multiple comparisons.
I have ran the analysis in a pair-wise mode. DESeq than gives this output:
estimating size factors # workflow of the DESeq2 function pair-wise modus estimating dispersions same number of samples and coefficients to fit, estimating dispersion by treating samples as replicates gene-wise dispersion estimates mean-dispersion relationship final dispersion estimates fitting model and testing
But when doing the DESeq() command with all the samples in one go, I sort of trick DESeq to think there are multiple replica in the data set and I get this output:
estimating size factors #multiple comparisons in one count table estimating dispersions gene-wise dispersion estimates mean-dispersion relationship final dispersion estimates fitting model and testing
is it better in this case to run DESeq in a pair-wise modus?
Don't I sort of cheat the results, when letting DESeq thinks there are multiple replica in the data set?
thanks
Assa
Can you show the design in both cases so we know what you mean by tricking DESeq() in the second instance? I'm assuming you have some sort of factorial design in the latter case.
This is the script I am using:
This is how I do it here, But I don't get the message, that DESeq2 "discover", that i don't have any replica. which made me wonder whether or not this is correct. In the first run, I always extract the columns I needed for the comparison, which was a bit tedious.
So when you do the "pair-wise" method, am I correct in assuming that you're just loading two samples at once? It's getting the replicate information from the treatment column, so it's correct in the second instance that you do in fact have replicates (at least, you're telling it that you do). You don't need replicates for everything, just one group.
Whether your design in the second instance is correct or not depends on the underlying biology, though since you're treating things as unreplicated in one instance, I would guess that this doesn't actually match the experiment.
Yes, I do have one replica for each condition (or two for some). The different experiments are independent of each other. This is why I first tried to run them separately.
So if I understand it correctly, I can't fit all the samples in DESeq in one go and than just extract the differentially expressed genes, if the conditions are not somehow related to each other. If the experiments were done as single, separate cases, so should be the analysis.
Thanks Devon