Hi guys,
I have a question concerning RNA_Seq and PCR analysis results. We performed an experiment in which we interfered with the expression of one gene of interest in order to change the effect on the functionally correlated genes. We performed the experiment using 2 replicates: 2 for treatment (interference) and 2 for control (not interfered gene). Through the PCR we verified that the gene was interfered properly and hence not expressed and this resulted with respect to both controls. Then we performed the RNA_Seq analysis on the full transcriptome but surprisingly there is not a statistically significant difference between the expression of the gene of interest between the treatment and the control although the cell lines were not changed between PCR and RNA_Seq, i.e. the samples were the same. We expected a confirmation of the PCR experiment.
Can anyone help me to explain such a strange behaviour between the two experiments?
Kind regards
B.
There are so many potential sources of error in both experiments that it is difficult to know where to start with an answer. How many house keeping genes did you use to normalize your qPCR? What primers did you use?
To be honest, this isn't something someone without a deeper understanding of your project could help you with. 2 replicates for RNA-Seq may not be enough to detect differences in lowly-expressed genes...
@John is right, there are many potential errors in the experiment.
How did you 'interfere' with gene expression? Was it a transient siRNA? A stable knockdown? A knockout? Stable shRNAs can be silenced over time. Knockouts (depending on the target site) can still be transcribed but yield non-functional mRNA.
What degree of downregulation did you get by qPCR? 20%? 50%? 90%?
The interference was done with transient siRNA by a technician in my lab. She observed ~2 fold decrease with the control (scramble).
Can you tell me how you know that:
Knockouts (depending on the target site) can still be transcribed but yield non-functional mRNA.
I am trying to find a paper but I can't:(
If you introduce a frameshift/nonsense mutation your RNA can still be transcribed but lead to nonsense mediated mRNA decay (NMD)
How did you do the RNA-Seq analysis? Do you have any numbers and units of measure?
Concerning the RNA Seq analysis, I followed the "best practice guidelines". Specifically (and briefly), I performed the alignment, the merge of the lines, I removed the duplicated, the sorting and I performed the reads count using "htseq-count -t exon -i gene_id" . Then I used the reads counts to perform the differential expression analysis using DESeq.
Remove duplicates is for calling SNPs in genomic data. In quantitative RNA-Seq you are removing more signal. Most read-duplicate software I know of looks at just the start and end coordinate, and assumes replications come from amplification. The best practices are for genomic variant calling. I think you want to count reads including duplicates to get everything that was sequenced. Your alignment parameters will also influence how many reads are seen at the gene, but hopefully for both samples in the same way.