I am preparing to perform a differential expression analysis on a count matrix of Human data that was quantified with featureCounts using Gencode v43 as the gene annotation, which means I have separate counts for the X-chromosome and Y-chromosome copies of the genes in the pseudoautosomal regions (PAR) of those chromosomes. Given that these genes generally behave like autosomal genes, it seems that if I want to treat them as such, I should sum the counts from both the X and Y chromosome copies of these PAR genes. This will result in a count for each PAR gene that represents the total count of reads from both alleles, which is the same as the count for each autosomal gene. For example, if a sample has a count of 10 for ENSG00000228572.7_PAR_Y and 5 for ENSG00000228572.7, I would change the count for ENSG00000228572.7 to 15 and remove ENSG00000228572.7_PAR_Y from the count table entirely. Is this the right approach, or is there a compelling reason to keep them separate?

Note that sex/gender is not generally of interest for this study, other than controlling for sex differences by adding a sex term to the DE model.