Dear every body you can use Pfaffl method that involve using their data when the efficiencies of GOI and Ref. gene are not the same

thanks

LinRegPCR really works well and I usually use it to test eff. after my qPCRs. As long as they are between 1.8-2 I don't really bother to calculate with individual efficiencies though. If eff for a gene is below 1.8 it is better to design a new primer set anyhow.

Did I get you correctly here: You have standards, and these standards are amplified with a different efficiency than the samples? If this is so then you do have some serious problem with the standards or with the samples (more likely with the latter). Try different purification methods. When targets are not amplified with alomost-optimum efficiency, the error in the final result can easily get very, very large, and an erroneous quantification result is not neccessarily recognizeable (for instance by large variance).

Just a number example (calculation details upon request): the variation in efficiency from 1.9 to 1.8 easily shifts the ct-value from 27.5 to 30, so by 2.5 cycles. If the efficiency of a given sample/PCR is detemined with a precision of plus/minus 0.05 (what is quite tough to achieve!), then the precision of the estimated result is roughly more than a factor of 4! If you dont care about factor 4, why bother with efficiency correction at all? But if you need more precise results, relying on sub-optimum efficiencies that are estimated from individual curves or from (external) dilution series (hoping that they are constant and similar between samples/PCRs) is no reliable way to go.

Thanks for your replies. No, I don't have different efficiencies in standards and samples, but in reference genes and target genes, which is why I can't use ddCT method. For the standards: actually I had a problem of degradation (CT-values of my plasmid standards slowly increased with the time), which I noticed, when I wanted to analyse the whole data set with the standard curve method. Therefore I didn't trust my standards any more and I started to search for alternatives to the standard curve method. On the other hand I don't have the degradation symptoms with my control samples, which I also run in addition to the standards on each plate. That's why I don't think it's a problem of sample purification or RT-PCR but of the buffer, in which I soluted the standards. I do trust my sample data and I think using the LinRegPCR method circumvents the standard and efficiency problems, as I can do relative quantification with starting concentrations directy achieved from my individual sample data.

If you have the same efficiencieds in standards and samples (per target sequence), then the standard curve method is perfectly fine (given you dont care *why* the efficiency of at least noe of your targets is sub-optimal!). The degradation does not chance the efficiency. As long as this is not too strong, using an additional calibrator is also the best way to correct the the sample ct's so that they can be analyzed with a given standard curve.

Note that LinReg does not avoid these problems. It exchanges more precise efficiency estimates from dilution series by less precise estimates from individual amplification curves. AFAIK there is no proper error propagation from the uncertainty in the efficiency estimation to the results.

qPCR with sub-optimum efficiencies remains *very* problematic, whatever software you use, because both, precision and accuracy, of the efficiency estimates are too low to obtain reliable results. Ignoring the uncertainty in these estimates, the results might look perfectly fine and have a very low variance - but they still may be considerably wrong.

I don't think my efficiencies are generally too low. Depending on the primer pair they are between 1,79 and 1,98. I just said, that they differ from gene to gene.

Another reason I don't want to use the standard curve method ist, that I want to compare the gene expression of my different genes (gene familiy) in a variety of plant tissues. If I use the standard curve method, I cannot actually compare results of different genes, because they are obtained on the basis of different standard curves (different efficiencies). Calculating the efficiencies per sample makes it possible to compare the calculated and not estimated results of each gene in one sample directly.

You criticise the lower precision of individual calculated efficiencies in LinRegPCR. The current version calculates a mean of all individual efficiencies and uses an improved the algorithm, so that in the end you should come out with similar precise result as you would with standard curves. Also the argumentation goes like ranking accuracy higher then pricision, which I think is a good idea.

Given this, is there any other reason, why you wouldn't use this method?

Nice to hear about "improved algorithm". Roche did something similar in the LightCycler software. The real reason was that the "2nd derivative maximum" method fails (too often with real data; in theory it's all fine). Affymetrix did it with their Microarray software. The real reson here: the PM/MM system fails (again, too often with real data; in theory it's all fine).

Maybe I was not clear enough: even from very carefully measured and large dilution series, both, precision and accuracy of the efficiency estimates are actually too low for quantifications that should be able to detect differential expression of about a factor 2 (or even more). One of several reasons for the limited accuracy are background drifts in the amplification curves. They are usually corrected, but the available correction methods are quite "empirical" and imperfect. It happens quite frequently that people determine values of >2.0 (say 2.2 or so). In this cases, they attribute the "overshoot" to "error" and give it a final 2.0 efficiency. Then, how about getting 2.0 or 1.8 right away? Is there also some bias? And if so, how big is it?

And again, using the slope of the log-linear segments of the amplification curves is a bad idea, since this part of the curve is (1) not well defined (start, end?) and (2) does not truthfully reflect the amplification kinetics in the cycles before, because here already considerable amounts of amplicons and other products accumulated. These are no such ideal conditions as in the cycles before (note that the amount only 3 cycles before is only about one eighth; 3 cycles after the ct, the amount is eight times higher. Finally, data from 3 - 5 cycles are used to extrapolate over a range of 20-30 cycles! This is just flawed.

I do not understand why a comparison of different genes using external standards should be impossible? This procedure actually *eliminates* the influence of the potentially different efficiencies (without any extrapolation!).

Thanks for your opinion! If you are interested in their algorith you should look it up in Ruijters et al, 2009 (probably you already did?). Basicly the idea is, that the qPCR amplification curve should give a straight line until the beginning of the plateau phase, which is simply overlayd by background noise in the beginning of the reaction. They determine the beginning of the plateu with the SDM and calculate the baseline backwards trying to get a straight line on as many points as possible. Anyway, I see your point, that probably there might be some influences by the already accumulated reaction products in the visible log linear phase of the reaction.

On the other hand, there might be a lot of factors which might lead to incorrect calulation of your standard curve efficiency (e.g. I also had standard curves with efficiencies of like 2,05 and I found it kind of awkward), so I assume you always have to compromise at some point.

As for the comparison of standard curve derived starting concentrations of different genes and why I learned I shouldn't do that - I suppose it is exactly because of that: you calculate all your data per gene with a standard curve efficiency, which might be error-prone to some degree and this assumed error might differ for each gene - so how could you compare them in a resilient way? If I put up with the systematic error of an algorithm like LinReg, isn't this more consistent in the end, especially if I use the mean from a lot of individually calculated efficiencies?

Best regard!

Given the efficiency (of some target gene) is the same in both, standards and samples, than their ct values are directly comparable. As you know the concentration of the standard, you can directly calculate the concentration in your samples by their ct values. Calibration curves are generally used whenever and wherever the analyte of interest is not directly accessible (many fields in physics, chemisty, biochemistry and biology - ELIZA as just one example). I am still puzzled why this should not be applicable in qPCR. The best Standard will be a linearized plasmid, but a purified PCR product from "outer primers" will be also good enough. The standard DNA may even be produced with the same primers as for the qPCR, but taking "outer primers" will more faithfully mimic potential effects of the adjacent, outer sequences on the first-strand syntheses in the qPCR. The only known drawback of this method is that the amplified products used as standard may cause contaminations. I recommend to prepare the standards in a different lab and only take the working dilutions back to the place where the qPCRs are pipetted.

Well, thanks for that statement, I will definitly think about it and I would also be thankful for other opinions.

But besides this gene-gene comparison question I still do have this degradation in my plasmid standards and I don't see, how I could use standards like this as calibrator samples for the standard curve. Maybe their efficiency has not changed, but their concentration obviously has. I though about adjusting their measured Ct-value in the last runs to the original Ct-values in the beginning but this seemed sowehow crude to me.

Anyway, I am still in the process of analysing and I will try out different methods for sure. I still favour the LinRegPCR method. I think it is important to be aware of the pitfalls it might have but I personally think it's quite useful, at least I will try out and compare. Thanks a lot for giving me feedback!

In my experience LinRegPCR gives a very good average of eff for a given primer pair, so if you have enough reps the method works well

I tried alot to download this software, but i could not. plz share the link from where it can be downloaded or email me the software application so i can install this software at [email protected]

Dear every body you can use Pfaffl method that involve using their data when the efficiencies of GOI and Ref. gene are not the same

thanks

You should include qPCR fficiency for every primer-sample combination in your relative gene expression analysis. LinRegPCR is very useful as it provides a way to calculate this efficiency post-qPCR. Then, you incorporate the resulting Efficiency Value in the Pfaffl (2001) method.