Hi, 

So I have the following scenario: I have a bulk RNA-Seq matrix with three conditions and three biological replicates per condition. I have used CIBERSORT to estimate the cellular composition of my sample, so now I have this value as well. Now I want to find differentially expressed genes (DEGs) for each cell population given its cell proportion. 

How would you recommend to go about with this using DESeq2? 

I had a suggestion to multiply the matrix by the mean of each cellular proportion and run DESeq2 on this new matrix. But this just seems too easy and wrong to be true. 

Any help is welcome! 

Although you have cell type proportions, you don't have the gene expression split among those cell types. Sometimes this splitting of the actual expression is called "deconvolution" (other times, "deconvolution" is used to mean, estimating the proportions alone). Just splitting the expression of each gene equally is not going to give you a satisfactory result. 

To think about: if this were easy to do statistically / computationally, we wouldn't need to ever do flow sorting (or to some extent single cell), once we've figured out the cell type markers / profiles. But it's a really hard statistical problem, and not necessarily identifiable.