qPCR reference gene testing help - (Oct/24/2013 )
I have a couple of questions about designing experiments to check my reference genes. In my lab they're not very concerned with this but having read the literature I think it is important. In my previous lab I did this using Taqman assay and relative quantification. Now, in this lab, we use SYBR green assays and the absolute quantification method. Is testing for the most stable reference gene the same with absolute quantification as with relative quantification?
My specific questions are:
1) For the standards we use plasmids, but having read some posts on this forum I'm not sure this is the best approach since it doesn't compensate for the efficiency of reverse transcription. Is there a better option for the standards? What are the drawbacks of this better option? How would one measure the RT efficiency and is there a way to incorporate this into the quantification?
2) I know what is required for each quantification method (standards, ref genes, calibrator...) but from everything I've read the following is still unclear to me: is it possible to do absolute quantification with the Taqman assay and relative quanitifcation with the SYBR green assay, and vice versa?
3) Do I have to run a reference gene on every plate? Or is it ok if, for example, the ref gene is on plate number 1 and I compare target genes on plates number 1-5 based on this reference gene?
4) I have RNA samples from 3 different conditions, 2 wells per condition (so this would be the biological replicate, yes?), in 3 experiments (n=3). Can I run all the samples from the 3 experiments on one plate, or does it have to be 3 different plates?
5) About Normfinder, geNorm and Bestkeeper - do I enter raw data, so just Ct values, or does it have to be quantified? Is it possible to use these programs with absolute quantification?
6) For my pilot experiment where I measure the stability of reference genes across samples and experimental conditions I had 3 genes - b-actin, g-tubulin and b2-microglobulin, I had standards from plasmids for each of the three, and I had samples for experiment no.1. I averaged the Ct values of the triplicates, and then averaged the biological replicates. Is this ok? Is the difference in Ct value alone enough to tell me which gene is the most stable - where the gene with the least variable Ct value between all samples would be the most stable gene. Do I need to run this with samples from the other experiments?
Thanks! I know its a long post, but I'm swimming in this sea of qPCR literature all alone and any life-raft would be greatly appreciated!
Even when using absolute quantification.. you actually do normalize it to a reference gene, right? The question is if you only want to know a relative abundance of a certain gene (i.e. 3 times more that in control sample - "the calibrator") or you don't have a calibrator and just want the target gene/reference gene ratio to do.. whatever with it, but this is rarely used. Unless you want to compare number of copies of different genes together, or minimal residual disease, pathogen detection, this is not done regularly.
If you say you did the same with relative quantification calculations, it means you probably don't need absolute quant. at all and you can bypass the need for "absolute" standard samples then.
Anyway the selection of the most stable reference gene is absolutely the same in this case. (unless of course you have that special measeurement like amount of BCR-ABL to wt ABL, where you don't look for stable reference, but you have a firmly chosen one)
1) It doesn't. I've seen having aliquotes of viral RNA for the absolute quantification of RNA viruses, when they wanted to know exact copy number of them. But as I wrte earlier, maybe you don't need any absolute plasmid standard if you can go with relative quantification.
2) SYBR and Taqman are just methods of detection, each has pros and cons (Taqman can be duplexes, it's supposed to have increase specificity, but you can't actually directly check the specificity on melting as you can do with SYBR, also SYBR is cheaper), but you can surely do absolute or relative quantification with both. That is just a way how you interpret the results.
For rel. quant. you need calibrator, and reference gene, it's better to run a standard curve too to assess efficiency and calculate with it. For absolute quant you don't need reference, but you need absolute "copy numbered" (or at least concentration or something) standard everytime you run it. Also, theoretically you don't need reference, but practically absolute results are mostly normalized too, that means you need two absolute quantifications, one for target gene, one for reference gene, and standards for both, then you divide the values.
3) In rel quant, you need to run a calibrator sample on-plate with each sample that you want to compare. You don't need to run reference gene(s) and target gene(s) on same plates, cause they will be relativized anyway in the calculation. In absolute quant, you need the standards for respective gene on-plate, again don't need reference and target at the same plates, they will be assessed separately anyway.
4) Don't really get it.. you have two growth wells or plates of cells for each condition (that would be a sort of semi-biological replicates since for a full biological replicate I would imagine that you woudl just repeat the whole experiment, but I think that is usually called a bio replicate, yes) and those "3 experiments" mean technical replicates? Or three different experiments, repeated in different times (with two wells each)? Technical replicates are be run on a same plate, I'd even say they must. You have to fill technical replicates (i.e. from the Well #1) into one plate, but if there is enough space you can of course put there Well #2 as well. If not, you run it on different plate.
Due to the fact they are independently calculated anyway, it does not matter, inter-run variabilty is usually negligible.
5) Can't tell about the other two, but GeNorm requires "quantities" (delta-Ct values: Ct of calibrator - Ct of sample, calibrator is one of the samples you chose, doesn't really matter), it is described in the manual. GeNorm is this way compatible with absolute quant too, you just pick a calibrator.
6) I myself am not clear on the double averaging, especially in respect to what SD is then displayed in the graph, but yes, first you average the triplicates, then biological replicates. I would personally evaluated technical triplicate SDs is they are OK-notOK (too big), and if this passes, then average these biological replicate, calculate SD, that would be displayed in potential graph as a SD of biological replicates. I don't know if it's possible to incorporate the technical SD into the biological SD, maybe even not.
(when read it to the end.. just a summary.. I think you don't need to bother with absolute quantification and plasmid stuff at all, do a relative, and make dilution standards (dilute the most abundant sample to get efficiency data))
just a short addition:
in my geNorm manual it says: "the highest relative quantities for each gene are set to 1. These raw -not yet normalized- reference genes quantities are the required datat input for geNorm."
So the correct formula is E ^ (Cq-min - Cq) were E is the efficiency of your calibrator-gene (measured in series dilutions)
normalization for your gene of interest ist done after that.
Yes, sure, my mistake, E(efficiency) raised to the power of (Ct of calibrator - Ct of sample). But "2" instead of real measured efficiency can also be used.
But there is no "calibrator-gene", efficiency is a characteristic of an assay (when measured, the dilution series had to be done for each of the housekeeping genes that are compared) and calibrator is one of the samples run with each gene. Efficiency doesn't apply for a sample, only for an assay (combination of primers/probe in set conditions and programme).
i was totaly shure, that you knew this =) But i think its worth to add such notes even to old topics.
Ha you´re right, calibrator gene could be misunderstanded.
Atm the calculation of the calibration factor drives me crazy... trying to do this with Hellemans 2007 paper.