This is a test version of Biostars. For the public version, visit https://www.biostars.org.
Question about analysis design in DeSeq2

I have cells from 5 healthy patients and 5 disease patients. Cells from each patient were treated with control and treatment conditions.

I would like to do differential expression analysis via DeSeq2 to understand how healthy vs. disease state affects the differential expression of genes between the control and treatment conditions. Within this, I need to control for the patient being the same between each control and treatment condition pair.

I have successfully figured out how to use Deseq2 for a simple analysis (eg just looking at how disease/healthy affect differential expression in the treatment group ONLY as ~Disease), but how would I specify the analysis described above?

Thanks!

rna-seq deseq2
• 2,031 views • link
updated by caggtaagtat caggtaagtat • written by jillianrjaycox • jillianrjaycox

Section 4.1 of the EdgeR manual gives an example of disease with matched control samples. Yes, that's a different package, but DESeq2 will work exactly the same.

• 1 views • link
written by Devon Ryan Devon Ryan

Be aware that the section 4.1 example is for tumour vs normal comparisons nested within patients, not for the comparison of diseased and undiseased individuals where there is not an explicit matching between pairs of control individuals and diseased individuals

• 1 views • link
written by russhh russhh

It looks like OP has a nested design as well (treated and untreated cells in each patient), but yes the models won't be exactly the same.

• 1 views • link
written by Devon Ryan Devon Ryan

1 answer

I think you could also use something like this in your design parameter of the DESeq function

design = ~ Disease + Treatment + Individuum

and then you could check if there are big differences in your result tables with and without takeing into account the individuum.

• 0 views • link
written by caggtaagtat caggtaagtat

I don't think that will give OP the design they want. WRT design matrices, this will give an intercept column, a disease column, a treatment column and 9 individual columns. There should really be 8 individual columns: 4 for each disease state. Is it possible to provide an explicit design matrix to DESeq2 nowadays? [The contrast is also nontrivial because patients are nested within disease state]

• 1 views • link
written by russhh russhh

Log in to answer this question.

More posts like this