EdgeR analysis with different biological replicates numbers problem

Question

[ahmad.azd]

Hi There I am doing an edgeR analysis using RobiNA software. I have three types of samples (1,2 and 3) each type has multiple biological replicates. all the samples counts are included in one file. I am trying to find the differential expression genes between samples types 1 vs 3, and 2 vs 3. my question is when I change the number of the biological replicates of the sample type 1, the analysis results ( deferentially expressed genes ) of 2 vs 3 are different, even that nothing has been changed in the samples type 2 and 3. is this normal or I am doing something wrong. and it would be appreciated if someone can include an explanation for this.

Thanks

[Kirill Tsyganov]

edgeR uses replicates to estimates variability in your data. Your sources of variability are technical (usually small) and biological (usually much bigger). And what @WouterDeCoster is saying your estimate of variability will change by removing replicates, thereby altering downstream analysis.

[WouterDeCoster]

I'm not sure about RobiNA, but if you take out a sample from your analysis then that will influence your normalisation procedure, and as such also the differential expression analysis.

[ahmad.azd]

Thanks for replying RobiNA is an R based software and this analysis use the glm functionality. My problem here is how to explain this scientifically with evidence. because the samples( here samples type 1) that been removed are not included in the analysis ( 2 vs 3), but they are only found in the same file.

[WouterDeCoster]

if you take out a sample from your analysis then that will influence your normalisation procedure, and as such also the differential expression analysis.