Effect of Over-Represented Transcripts on DEG Analysis

[evanso08]

We have a fairly simple set of RNA-seq data that we are using to perform a DEG analysis, breifly, a wild-type strain and a single gene knock out (in yeast). One issue we have is that the yeast is auxotrophic, meaning our wild-type line does not produce transcripts of the gene TRP1, while our knock-out selection marker is TRP1, meaning there is a huge fold change of that gene in the DEG analysis. We are also noticing that there seems to an "over representation" of certain genes, representing a large portion of the sequenced reads. This seems to shrink the fold changes of genes we know are differentially expressed, verfied by RT-qPCR. Is there a way to account for this "diltution effect" when using a DEG program like deseq2?

[fchatonnet]

Dear evanso08, to deal with TRP1, if it's only expressed in one of your strains and you've verified that you find only transcripts of it in the mutant strain, just remove it from the analysis as huge fold changes can sometimes distort the fitted dispersion of other genes.
As far as over representation of other genes, if this phenomenon is comparable between your two compared experimental conditions, you may also just remove these genes from the analysis and perform a DESeq2 DEG identification on the remaining genes. That's common when studying cell types which produce large amount of specific RNA, like immunoglobulin light and heavy chains for B lymphocytes for example.
Finally, it could be interesting to investigate why you get such an over-representation and what is the function of the over-represented genes.
Good luck!