We sorted CD8+ effector memory cell based on the level of expression of a certain protein, p_high and p_low and obtained bulk seq data from these sorted cells. After running DESeq2 I realized that there are significant fold changes related to epithelial and fibroblast genes in the data that mainly seem to come from the "p_low" samples and my question is how to best deal with this contamination in the data?

For the genes in question I have double checked and they are exclusively expressed by epithelial cells or fibroblasts, not CD8+ T cells.

I have thought of a few things

• Defining the contamination as a covariate. However, given that it appears to be associated more with p_low, regressing out the contamination as a covariate would also potentially regress out biology from p_low.
• Simply removing the genes, however that would still leave uneven reads between samples that had more or less contamination.
• Removing the samples that have the most contamination. However that still leaves the other samples with contamination that could still affect the DESeq2 results.

Has anyone dealt with this before? Does anyone have advice on how to best approach analyzing this dataset?