I am analysing a single cell RNA seq dataset with 6 responders and 8 non responders and doing pseudo bulk differential expression between groups using edgeR. There is gender based confounding factor when dealing with it with females (n=3) only present in non responder group. I made a design matrix with gender+response and used in edgeR but DE still contains gender specific genes due to the female group only present in non response. Is there a way to deal with gender as a confounding variable?

You can exclude the females from the non-responders.