Thanks! I just took a look at whether those downregulated genes are enriched for genes with many zero counts across samples in each comparison group and that appears to explain part of it. (https://ibb.co/g4yd1hD) I suppose it could be that a gene is not expressed at all in group A, but is lowly expressed in group B, however in that low expression space, not sure how confident we can be in such a result.

So, I noticed that part of the inflation could be explained by contamination by another cell type (see here: https://ibb.co/RDmbpQ6). Based on this and the previous observation that a lot of these genes had zero counts in many samples per group, I proceeded to filter out genes related to contamination, and kept genes that had more than 5 counts in at least 3 samples per group. The MA plot looks much better now: https://ibb.co/YQTpSFS.