I am validating my primer which will be used for relative quantification by doing a serial dilution. For 2 of my dilutions I have an outlier that is affecting my R2. I know an acceptable R2 is anything above 0.99. After omitting my 2 outliers my R2 improves to 0.992 from 0.963, while the efficiency reminds about the same, at 86%. I want to know if this is acceptable to determine if my primer passes. Any help will be appreciated.
Hello, in my opinion, you can omit the 2 outliers. When looking at your picture it is clear that you have two replicates that are nearly identical and that the outliers are most certainly do to pipetting error. They occur and induce a loading of more or less template DNA in the reaction creating this difference. Of course, it would be better not to have them but they happen sometimes and they are the reason why we all do replicates of samples (dilution standard and samples analysed). If you really have a doubt, you can redo the standard just to be sure. I would advise to put the standard each time you will use this primer pair, it would be more robust when analysing the data of your samples. Hope my answer helped.
Good luck for your experiments.
I think there is significant issue with std curve. Repeat experiments have more than 1 ct shift. If the std curve repeats are not reproducible, then it will affect quantification in test samples. The efficiency (0.86) also looks low, typically for qPCR it has to be above 90%.
I use Dixon's Q test for this sort of calculation. The wiki entry will show you how it works, and includes the primary citations you will need.
http://en.wikipedia.org/wiki/Dixon's_Q_test
I also think than being picky on "outliers" is a silly thing to do here.
0) Forget this R2 - it does not tell you anything useful in this context.
1) There seem to be only 3 points per concentration. This is nothing. Well, they scatter. Seeing on accidentally being farer away from the rest also means seeing 2 being accidentally close together. No reason to suspect outliers.
2) Having only 4 concentrations is again nothing. How will one judge linearity? Fron the diagram one can argue as well (!!) that the measurements of the highest conc. do not fit, that these are the actual "outliers". In this case the slope would really change (-> steeper -> even lower efficiency...!)
3) Declaring 2 points of 12 as "not fitting" or outlying means to discard more than 15% of your data. If so many values can be "bad" already in your beautiful standards, how can you believe that the ct values you get for your unknowns will be even slightly reliable?
So you have three plus one issues:
- bad design to determine the calibrtion curve. It would be better to measure more values over all and to distribute them better over the dynamic range (i.e. take more different concentrations; do a proper residual analysis)
- the whole data does not seem to scatter around a straight line
- the efficiency of below 0.95 (and 0.86 is considerably lower!) indicates either a suboptimal amplification of suboptimal analysis of the signals. When the efficiency is so bad, I would not trust that it is constant during the whole PCR and that it is the same for all unknowns (tiny differences in the samples may alter this fragile system). And when the signal processing is bad, well...
All these issues are connected. A bad design makes it difficult to judge linearity. Missing linearity can indicate a problem with the signal processing and so impact the determination of the efficiency.
The additional issue is the large variance in the ct values you already see. Actually your ct's should not vary by more than plusminus 0.15 cycles. If you had measured 200 cts, it would be tolerable to get one or two "outliers".
- Badges
- Science method