Dear all,

I have an experimental setup with one control sample and two knockdown samples done in two replicates. I made the two knockdown samples using two different siRNA construct so my problems is that i don't know how my deseq design should look like if i want to take into account both the differences between knockdown and control, but also between replicates and siRNA constructs. One note is that one of the knockdown samples were of poor quality and were not sequenced so i have 2 controls and 3 knockdown samples. I have tried the following coldata:

rownames     Condition     Replicate     Sample     

CTR               ctr                a                   c

CTR               ctr                b                   c

KD                 kd                a                    d

KD                 kd                b                    d

KD                 kd                a                    e

I have tried every possible combination for design but also for contrast when getting the results so i think that the problem lies in the structure of the coldata.

I hope you understand my question and thank you in advance

-Kristian

A fixed effects model like DESeq2 cannot control for the sample differences and allow you to compare KD vs CTR, because these are confounded. You can instead use limma-voom with duplicateCorrelation().