sRNAseq with NGS can tell you something about the relative abundance of specific miRNAs or other molecules compared to the total small RNA transcriptome. However, you will not be able to make any absolute quantitative statements from NGS data.

Seeming upregulation of a set of miRNAs may also come from downregulation of other miRNAs. In fact, all miRNAs may have lower/higher absolute abundance compared to the control. This is something that you cannot conclude from NGS.

In other words, using NGS data you compare the read counts of a specific miRNA relative to all mapped reads. With RT-qPCR you compare expression of a specific miRNA with expression of one or more reference genes. That often makes a difference, particularly when the total amount of small RNAs compared to the total amount of mRNA changes between different conditions.

sRNAseq with NGS can tell you something about the relative abundance of specific miRNAs or other molecules compared to the total small RNA transcriptome. However, you will not be able to make any absolute quantitative statements from NGS data.

Seeming upregulation of a set of miRNAs may also come from downregulation of other miRNAs. In fact, all miRNAs may have lower/higher absolute abundance compared to the control. This is something that you cannot conclude from NGS.

In other words, using NGS data you compare the read counts of a specific miRNA relative to all mapped reads. With RT-qPCR you compare expression of a specific miRNA with expression of one or more reference genes. That often makes a difference, particularly when the total amount of small RNAs compared to the total amount of mRNA changes between different conditions.

How many genes did you validate?

Short answer: If NGS and qPCR results overlap (i.e. same direction of expression): great news! If not, I'd always trust the qPCR (if performed right) results over the NGS.

Before I give you the 'Long answer', and to develop a more coherent hypothesis on why you have this discrepancy, It'd be useful to know (as David partly mentioned) if your NGS (and qPCR) data are intended for relative (i.e. Treated vs Untreated, Disease vs Normal, Phenotype 1 vs Phenotype 2, etc) or absolute quantification (i.e. counting how many copies of that specific miRNA [or gene] in a specific phenotype). I'd assume (hopefully) you are in the former situation.

Long answer:

When you use a technique/technology (e.g. qPCR) to validate another technique/technology (e.g. NGS or Microarrays) you would always trust in the one you are using as a validator (qPCR in that case).

Still, you should note that analyzing the 'same samples' with another technique is just a 'technical validation'. That's, you validate the observations you saw with NGS with a more sensitive/accurate and low-throughput method (like qPCR). As David mentioned, you are getting only a relative quantification (i.e. the expression of your gene(s) of interest in a certain phenotype/condition relative to its expression in another phenotype/condition (i.e. control) from both techniques, and I guess you are not interested in knowing the difference between absolute vs relative quantification! You are interested to know why you have this discrepancy (in terms of gene expression direction/trend) between the two techniques.

If you are comparing one phenotype/condition to another, I'd recommend the following:

1- Double-check you are using the right sample(s) as your control. If you use (for example) Treated vs Control in the NGS and you invert it (Control vs Treated) in the qPCR analysis, there is no doubt why you are getting opposite results.

2- 'Biological validation': try checking the trend of expression of this miRNA in another set of samples that belong to the same condition you are testing (not the ones you used in the NGS analysis) and check the direction of its expression. If you get the same results (with qPCR), then you should re-check your NGS analysis from scratch.

Hope this helps

An easy reason for this discrepancy would be primer specificity, is it really targeting the miRNA of interest?

Another reason would be instability of the reference gene(s). How did you do your reference gene study?

Thank you for your answers David and Ramy, I'm going to repeat the qPCR using different housekeepers as I'm unhappy with their stability. We're also reviewing the bioinformatics, to date to ensure these are correct. I think it's best to ensure that our experiment is technically sound in all aspects before we look for a biological explanation. I also take on board the fact that the read counts are normalized by mapping all read counts to miRNAs in the NGS and relative quantification is dependent on the housekeeper used in the qPCR as David mentioned, so will follow this, if our repeated experiment yields the same results.

Marc Paul O'Sullivan Hi Marc

I appreciate that if you explain how you solved this problem

Hi Reyhaneh Ebrahimi Khaksefidi

I solved the problem by removing the housekeeper U6, this is a small nuclear gene that is not good for serum or plasma, but should be suitable for Tempus or other whole blood-based method.

My suggestions for anyone working on this is:

Thanks Marc Paul O'Sullivan , I used two different reference genes, their expression pattern is highly similar, however, completely different from NGS data.