RNA seq data shows no gene expression but qPCR shows the same gene is upregulated

Question

[ravihansa82]

Recently I did RNA sequencing using ovarian tissue samples to find out the differencial gene expression in ovary based on WT and mutant gene. Bioinformatics results analysis show one of the gene expression is 0. However, when I do qPCR for the same gene I can see upregulation of that gene in several folds and I confirm this results by repeating the qPCR. Now I am confusing, why RNA seq data show no expression of such gene and my qPCR show upregulation of the same gene. The given gene is having one transcript. Can anybody explain me why this is happening?

[shawn.w.foley]

A few questions:

1) Do you mean the gene is not expressed via RNA-seq, or that there is no fold change?

2) Have you done melt curves with the qPCR to confirm this is not off target?

3) What are the CTs for this gene via qPCR? Is this a very lowly expressed gene?

4) What is your sequencing depth for RNA-seq? How many replicates? Is the fold change zero, or is the change not reaching significance?

5) Is your qPCR using multiple housekeeping genes for normalization? Are you confident that the housekeeping genes are not changing?

[ravihansa82]

Hi Shawn, what I understood from the Bioinfomatician, he gave me a list of ensemble gene IDs with the number of read counts for indivudual genes. in that data there were many genes showed as 0 reads. in that 0 gene list I found one of the universal known gene read counts also 0. when I do qPCR for that gene, I got CTs at 18. in qPCR housekeeping genes are not changed. at the moment I do not know what is the Depth for RNA-Seq.

[ATpoint]

Zeros in bioinformatics/RNA-seq do not necessarily mean that a gene is not expressed. Rather it can be that the gene is of repetitive nucleotide composition and therefore difficult to map. Alignment software might have flagged reads as multimappers and therefore these reads might have been ignored during the quantification, resulting in the zeros (probably more likely for shorter genes). Can you show a selection of this table including the header? It can still be that, as suggested above, the gene is lowly expressed at a level below the RNA-seq detection limit.

[shawn.w.foley]

It sounds like you have a file with the number of reads mapping to each gene (or is it just a subset of genes?). Take that file and sum the columns to get an idea of how many reads you have mapping to any gene. ATpoint and Kristoffer both made good points about multimappers, but I'd still be curious how deep the sequencing is before making any conclusions. Without sufficient depth then you could be seeing a lot of false negatives.

Answer 1

[Kristoffer Vitting-Seerup]

In my expression false zero expression of genes happens most commonly if haplotypes are considered whereby multiple (almost) identical version of the genes will be quantified - and since old school (aka on the verge of obsolete) methods only counts uniquely mapped genes there will be zero counts.

As a quick sanity check get the bioinformatician to just extract reads that overlap the genomic region of interest to see if there are any. Should be very easy from the bamfile(s) via a command like this:

samtools view yourBamFile.bam chr1:3052830-3052840

Where the coordinates (chr : start - end ) should be swapped for your gene of interest. If this works the bioinformatician would need to requantify the data taking haplotypes into account and then you should also try convincing her/him to switch to more modern quantification ways - you can read more about them (and why to use them) here.

Cheers Kristoffer

[ravihansa82]

hi Kristoffer, I aligned the sequence reads to the target chromosome. finally, I found sequence reads align with the target chromosomal region. thanks a lot for your suggestion.