For 3 separate RNA preps like you describe those differences in Ct value are normal. In addition to the real variation between biological samples, the differences you see will be explained by RNA yield and quality and technical variables (like pipetting). Generally, people publishing qPCR data show the means and standard deviations of 3 technical replicates (where the same cDNA is analzed by qPCR in 3 different wells). This allows them to have very small error bars and statistical significance. They then indicate in the text that they performed the experiment 3 times and had similar results all three times. Ideally, they will show the most representative result, although showing the prettiest one might be tempting.

As I understood, for each single cDNA you performed PCR in triplicate. Therefore it is not due to variation between biological samples. And you loaded same concentration of cDNA in each three wells, so these much variation in Ct is not correct.

I recommend you to be sure about cDNA concentration and pipette it properly. In my case, I had this variation in CT which we solved it by using calibrated pipettes and changing the tips for each well to load same amount of cDNA in triplicate wells.

Good Luck

Thank you both Mark and Somayeh for your quick reply to my question. Sorry Somayeh I think I may not have explained my issue to the best of my ability. The triplicates are experimental (by this I mean they are triplicates of my Tissue culture experiment and not PCR triplicates). So for each time point I had three wells containing cells and bacteria. Mark, after hearing your reply I feel a little more confident in my data. Maybe I should repeat the PCR two more time so that I have the nice tight error bars. However, like you say this is not the true variation of the experiment and simply just a measure of how good my PCR pipetting is.

You have close to 4 Cts difference between your biological reps. That sounds a bit too much to me but I am not familiar with your system. What was the variability between technical reps? And most importantly, if you saw the same variability in your "reference" genes, perhaps what you are seeing is just the actual differences in input RNA/cDNA, as Mark pointed out. Hopefully you did all the proper QC checks for RNA quality.

Yes definitely check RNA quality. And I'm presuming you generated cDNA from the same concentration of RNA between triplicates? If not, that might explain why you may be also seeing the same amount of variation in Ct values with your reference gene. In that case, go ahead and calculate your fold change and you might be surprised to see that the induction/decrease in expression may be quite close between the triplicates. You might also check the threshold value for each qPCR experiment, that's if all triplicates were done during separate runs. Also, did you do all 3 experimental replicates at the same time? I've found that sometimes even though you repeat experiments the same way and theoretically that should yield very similar results, but sometimes you don't get that close precision that you hope for. There's always a little biological variation. That being said, I typically do at least 4 biological replicates in case I have an outlier.

@Mark: doing stats on technical replicates and publishing them as you describe is plainly wrong. You can do this only on the biol reps, that's why you do them. The way you describe will only give an indication of your pipetting skills, and not of the biological variation. You can't cherry pick the one you like. That said, I always thought people do it like you describe to get unbelievably small SEs...

Hell all and thank you for taking the time to answer my question!

RNA was extracted using Trizol, cleaned using ethanol precipitation and then the yeild checked using a nanodrop. Purity ranged between 260/280 1.89 -2.1 260/230 2.12 -2.29. So purity was fine. I use Promega's GoScript Reverse transcripase system that calls for 0.5ug mRNA as a template if using oligoDT15. So other than the actual variation in my biological replicate the only other place I feel could have introduced any sizable variation is the efficiency of the reverse transcription. Although I added 0.5ug template there may have been variation is reverse transcription reaction between samples at this stage.

Like you have suggested Catherine I have actually calculated the fold change for each single sample and then also as a triplicate average and the results are strickingly simialr. and for those that are a bit more out that I would have liked the expression pattern remains the same.

Oliver, Mark is right, I see papers time and time again that have such small error bars that it just makes you very sceptical; especially when you are dealing with cells and bacteria in the same culture. I definitely think people do report the error for technical reps and not for the biological reps, as they should.

Sorry I meant "hello Al"

Robin, it sounds like you are OK with your results - and apparently you also have stable reference (AKA housekeeping genes) since your 2 to the ddCT-calculated final gene expression patterns remain the same among your biological/experimental replicates. Running an inter-plate/inter-experimental calibrator each time would allow you to formulate an adjustment factor for your Cq values to get them all in the same scale, but, it seems you have already attained what you are looking for. Using the same machine threshold on a per-target basis would be important here (as mentioned above) to avoid introducing undue variability in your final analyses. I assume your technical replicates on each plate are very similar (for all Cq values

@Oliver: I agree with you….but it's difficult to be "right" with a rejected paper. You might as well see how the game is played and play it that way. Again, it all depends what the title of the figure is. If you want to say that XX cells infected with bacteria always express precisely 2ng/ul of mRNA YY, then biological replicates is perfect. If you just want to say that infection with bacteria leads to an increase in mRNA YY expression then it is ok to show just qPCR technical replicates with tiny error bars as long as you have done the experiment 3 times and in all 3 biological replicates you see the same result. For a cell line there is little excuse for variability between biological replicates, but if you're working with human donor samples, for example, you will see great variability from sample to sample and calculating means and standard deviations of these biological replicates will lead to huge error bars.

@Mark, I think it would be better not to show error bars because n=1 and I assume you don't want to say with your figure you are able to pipette accurately. This also means you can't do statistics, so you can't test if the results are statistically significant ? Would probably be better to show the three points in the figure.

I totally agree with Jack, in this kind of experiment you cannot rely on the Ct values, you need a reference (Housekeeping gene) and then work with the DDCt. This should avoid errors introduced by pipetting and/or reverse transcription.