Dear Developers
We are trying to perform differential expression in a conditional knockdowns samples where the functional domains are excised using inducible CRISPR system (ours is parasite genome). Our qPCR shows depletion of protein expression. However when we do RNASeq, the gene is not deferentially expressed. When viewed in IGV, we see high coverage in the UTRs and adjacent exons both at 5 prime and 3 prime. When we remove UTRs from all the genes in the GFF file, the gene becomes differential (down regulated as expected). Is this a valid approach? We couldn't find any paper in the literature that used this strategy? The genes we are working are hypothetical and we try to check for alternate isoform, but the alternative transcript (which would code for a shorter protein than what the canonical transcript) was absent when we performed CoIP experiment.
Another strategy we tested is to limit the read counts to the CDS region rather than exon because we were measuring protein abundance using qPCR (after reading this article where it has been explained than entire exon is not transcribed into protein) and since the qPCR results weren't matching with RNASeq we though to perform this exercise (our first and last CDS is 300-700 base shorter than the first and last exon, regions where we see a lot of gene coverage and that's inflating the read counts of gene of interest ). But is this valid strategy too?
I would like to gather your thoughts on this given your years of experience with the RNASeq data.
Preprocessing that was done was
- Adapter trimming. reads shorter than 75 bp were removed a minimum quality threshold of PHRED score 20 was applied.
- rRNA removal using ribodetector
- gDNA assessment using gDNAx package. No contamination was found.
- fastQScreen to check for other contamination. The alignment reads were also checked. MAPQ was >= 60.
- No gene duplication in the genome.
Consider asking at biostars or bioinformatics reddit, since this is not directly related to a Bioconductor problem.
Anyway, a couple of thoughts:
Mind terminology. qPCR measures transcripts and as such does not measure protein abundance. Why not simply doing a Western blot and if that shows depletion just leave it be? If functional knockdown is confirmed the RNA-seq does not need to confirm it since the validation was already done.
Why not simply showing a screenshot of the genome brwoser asa supplemental figure? Stats are just one way of showing evidence for differential expression. If you make a knockdown and confirm by qPCR and show coverage-based evidence that would be perfectly fine for me as a reviewer. I would much more focus on downstream analysis that forcing your counting to perfectly validate what was already validated by qPCR.
This article, frankly, just points out bona fide knowledge that every serious researcher in this field should have. Of course exons of non-coding genes are not translated, that after all is the definition of such a gene, and of course unTRANSLATED regions are not translated -- they're regulatory and not coding.
In a nutshell, my advise is to focus on downstream analysis and present the qPCR + IGV (and maybe an additional Western blot) evidence as knockdown validation rather than spending too much time here.