Hi,

I was wondering if someone could provide some advice on the best way to approach my differential expression problem, or if it is even possible to do so.

We generated RNA-seq data from a series of gene knockdowns in an insect cell line:

I would like to discover differentially expressed genes for every Target vs Ctrl. Specifically, we want to know how similar the differentially expressed genes of Targets B-E are to Target A's DE genes. (We suspect A interacts with B-E and would like to see how loss of these factors compares with A).

However, the two experiment sets (batch A and B) were performed several weeks apart, and PCA and distance analysis shows Batch B is quite different from both the Ctrl and Target_A samples.

I attempted to use DESeq2 with a design incorporating Batch, but I get an error about the design not being full rank. I don't know what I'm doing with regards to design specification very well.

Do I have any options besides design = ~ Sample when setting up my DESeq object to account for batch effects here? Or are there potentially other options to consider?

Thanks,

Michael Chambers

hi Michael,

The experimental design is problematic here if you are interested in comparing the differences between B-E over Ctrl and A over Ctrl, because the batch variable is confounding this comparison. How could one know if the difference is due to B-E or due to batch B?