More up-regulated genes annotate to known GO/KEGG than down-regulated genes

Question

[bioinfo_cat]

Hello,

I have a dataset of RNA-seq data where I needed to do a de novo assembly. I am curious if anyone has any ideas why my dataset has an overrepresentation of DOWN regulated genes when I calculate DEGs using all of the gene count data (even the genes not annotated using blast) but when I use the genes that annotated to Swissprot/GO/KEGG/PFAM there is suddenly an overrepresentation of UP regulated genes?

They are almost completely opposite. Has anyone seen this before?