I have raw qPCR data. 2 samples "1 control and 1 with gene knock out". I made a serial dilution of each sample and performed a qPCR using 2 reference genes and 3 genes of interest. Now I have the raw data “the CT and average CT of each sample. I want to present the data as a chart. What exactly do I put in the chart. The delta delta CT? Or the 2^-delta delta CT? Or something else?
and do I put all the dilutions in the chart? or just the undiluted original sample? or calculate an average or a geomean of the sample and the diluted samples?
Another question. When I have more than one house keeping gene or reference gene, can I take the geomean of the average CT of both genes to calculate the delta CT?
...why did you do a serial dilution?
The only reason to do a serial dilution is to establish the range over which your data is linear (cDNA too concentrated: inefficient PCR. cDNA too dilute: target too rare for accurate quantification).
Once you've established that (and it's usually a generous range), you just...use that.
I dilute all my cDNA 1/20, and use 2ul per well. I do that for basically everything, regardless of target abundance, because this dilution is sufficient to allow PCR to proceed efficiently and quantify even super abundant targets (like 18S) while also not so dilute I can't pick up rare targets.
Beyond that, for data plotting (assuming you pick a single dilution and just use that), I would suggest just plotting dCt values. Or, if you like, -dCt values (since then increases in expression have higher values): these are normalised (because you've normalised them) and are also normally distributed, because they're still in log space. If you convert them to linear data (2^dCt) you now have lognormally distributed data, which is more challenging to present and also not as fun for stats.
Don't use ddCt values, because this is a very 'lossy' format: you take data where you can present individual sample expression values and make it instantly clear to the reader how consistently up/down expression is between groups, and boil them all down to a single fold change value.
For normalising, use the average of the Ct values of your multiple references. As long as they're still in logspace (Ct, not 2^Ct) you can use geometric mean or arithmetic mean: both will behave similarly.
John Hildyard
Thank you so much for you answer.
I just want to clarify that the purpose of the experiment is to validate the KO of the gene. So, I designed a primer that amplifies a region that should be deleted in the KO sample. and another primer around the loxP site. I guess my boss asked me to do a serial dilution to test the efficiency of the designed primers "please correct me if I am wrong and if this does not make sense".
I have tried using the average of the CT of both reference genes to calculate dCt, I then calculated the average dCt of the control samples "the control and its dilutions". Then calculated the ddCt using the average of the control. and finally calculated the 2^ddCt. The resulting numbers for the control and its dilutions were around 1. and if I calculate the geomen or the average of the 2^ddCt of the control it equals 1. I then also calculated the geomean or the average of the KO sample and its dilutions and the result was around 0.47
I also tried the Pfaffl method where I calculated the RQ - geomean of reg genes RQ - RQ goi/ RQ geomean of ref gen. Then gain calculated the geomean of the results of each group (control and KO) and I got the same results. Control = 1 and KO = 0.47
Is what I did right? Can I conclude that the goi or the segment of DNA between the primers is almost a half less expressed in the KO sample relative to the control? and just plot this on a bar chart?
It might be helpful if you shared some raw data: it would be easier to troubleshoot. I'm not...entirely sure what you're doing, from the description (sorry).
I also don't quite understand why a knockout would result in only a modest reduction, but that might be your model system?
Anyway, if you're happy to share some raw Cq values (can anonymise everything to "GOI #1" and "Ref#1" etc, if desired), we can work through this.
John Hildyard
Thank you so much again for your help.
I would love to share some raw data.
- my Ref genes are: GAPDH and PPIA. my GOI lets call them GOI#1 and GOI#2.
- I had 2 samples: control and KO. each sample was diluted by half, so I had sample 1, sample 1 (1/2), sample 1 (1/4), ... and so on. And same for sample 2.
For simplicity lets only take into account the original sample (not the dilutions). and lets take GOI#1 as an example.
Average CT of sample 1 (control):
GAPDH = 15.86
PPIA = 20.82
GOI#1 = 22.04
Average CT of sample 2 (KO):
GAPDH = 15.42
PPIA = 20.53
GOI#1 = 23.09
First step I calculated the mean Ct of the ref genes (GAPDH and PPIA):
for sample 1 (control) = 18.34
for sample 2 (KO) = 17.98
Then using these average Cts, I calculated dCt for each sample:
dCt sample 1 (control) = 3.7
dCt sample 2 (KO) = 5.11
Then I calculated ddCt for each sample:
ddCt sample 1 (control) = 0
ddCt sample 2 (KO)= 1.41
Finally, I calculated 2^-ddCt:
2^-ddCt sample 1 (control) = 1
2^-ddCt sample 2 (KO) = 0.37
does that mean that the GOI is less expressed in the KO (37%) compared to the control?
am I doing it right?
Yep, your math checks out!
It's...a bit convoluted, but it gets you to the right place. You can save time after the dCt step by just directly subtracting:
3.7 - 5.11 = -1.415
2^-1.415 = 0.37
But yeah: your method in principle is solid.
I would report this as "~40% of WT levels", because excessive precision is risky.
Having said all that, comparing two samples is even more risky: even if you only have "one control, one knockout" (assuming, say, this is a cell line), I would still recommend you do at least 3, and ideally 5, biological replicates, such that you can plot replicate dCt values and get a better handle of the range of that "40%".