Minimal TPM threshold for qPCR measure of gene expression
1
0
Entering edit mode
3.9 years ago
tlorin ▴ 330

Hello,

I am not a molecular biologist and I am mostly doing RNA-Seq. I want to find marker genes based on RNA-Seq results and I have to discuss with people at the bench about candidate genes would be good markers. These genes should thus have a measurable expression using qPCR.

I am wondering what would be the minimal TPM I should take to be able to detect the expression of a gene using qPCR.

In this paper it is written:

First, genes were filtered based on a minimal expression of 0.1 TPM in all samples and replicates, to avoid the bias for low expressed genes.

Does this mean that all genes with a TPM over 0.1 would be detected by qPCR? Does anyone have the experience of crossing RNA-Seq and qPCR results?

Many thanks!

PS: My TPM values are coming from the tximport function after kallisto pseudo-alignment.

qpcr RNA-Seq TPM kallisto tximport • 2.8k views
1
Entering edit mode
3.9 years ago
Charles Plessy ★ 2.7k

For the sake of easy calculation, let's imagine that:

• Detection threshold for PCR is 10 molecules (90 % loss);
• Reverse-transcription is 1 % efficient (99% loss);

Then to detect an expression level of 0.1 transcript per millon, one would need to load 1010 molecules in the RT-PCR.

Let's imagine further that:

• All RNAs are 1000 nucleotide-long;
• Total RNA is 1 % mRNA (99 % loss);

Then, one would need 1015 nucleotides of RNA in the reaction. This makes ~500 ng of total RNA. This is a very approximative estimate, but it suggests that attempting to lower the TPM bar will make some experiments unfeasible (collecting micrograms of RNA is a big challenge in some studies).

0
Entering edit mode

To be sure that I correctly understand, with a TPM threshold of 1 I would (roughly) be able to detect candidate gene expression with ~50 ng total RNA?

1
Entering edit mode

For the sake of easy calculation

That is exactly the point. This calculation is a nice theoretical thing, but please DO NOT!!! make any plannings/predictions based on this. RT efficiency greatly varies between kits and dependent on RNA quality/purity. Not all RNAs are 1000bp long and mRNA is not always 1% of total RNA, especially if your sample is not of best quality. Leave alone pipetting errors, RNase contamination and imprecisions in measuring RNA concentrations, even at 100% sample purity.

0
Entering edit mode

@ATPoint do you mean that the detection would greatly vary from one gene to another and from one experiment to another? (Sorry, I'm not familiar with molecular biology yet...). Hence, I presume the TPM would not "mean" much and a gene with a TPM of 0.1 could be detected by qPCR while a gene with a TPM of 10 could be not detected; am I correct?

0
Entering edit mode

All I say is that the reality is far to complex to allow predictions based on that calculation (which Charles most probably only posted to illustrate how difficult under optimal theoretical conditions such an experiment would be).

0
Entering edit mode

Yes, it is just an illustration that 0.1 TPM probably translates into "not a piece of cake" at the bench.

0
Entering edit mode

Then, which TPM would be a "piece of cake"? ;-) In other words, if you had to advise someone doing qPCR in your lab, would you consider a TPM of 2/5/10 as relevant for first test? (Keeping in mind that this will be highly dependent on the factors you mentioned above)

0
Entering edit mode

You wrote « I want to find marker genes » in your original question. This is a hard problem that can not be summarised in a few lines, and that can not be solved by just picking a bunch of genes for making presence/absence calls. I recommend you to seek advice from somebody who has experience in the field of biomarkers.