It's a problem called pseudoreplication - when you have multiple data points for the same trait taken from the same biological replicate (at the same time point). Although technical reps have an important role telling you how much variation arises from lab analysis, its important that you distinguish the two types in your statistical analysis (if technical variation is important).

I agree with Christopher, if the variability in your technical replicates is not important, then use the average of them, so that you have an average "technical rep" value for each of the 3 biological replicates. If you're not sure whether or not the variation between technical replicates is important, you could analyse it with an ANOVA. The ANOVA would have your 3 biological replicates, experimental factors (histone modification?), and also "technical replicate" as a factor. Hopefully technical replicate will not be a significant factor! If it is, then I another way could be a "nested" ANOVA, where you specify that each biological replicate has 3 technical replicates (this would recognise which 3 technical replicates came from each biological replicate. I hope that helps and makes sense.

Hi,

How many experimental conditions do you have? are running the qPCRs for several genes at the same time for same samples or the whole biological and technical replicates in the same running for each individual gene (it should be a better option than the former)? Why haven´t you apply three technical replicates per sample (because with two you may have always one high and other low!). If you have enough material you should repeat the qPCRs analysis maybe with a better previous design (I don´t know how it was maybe it was correct).

If you have no choice you could take into account the average of the whole biological and technical replicates as reference value for each group and try to eliminate the outliers (the least possible).

Other suggestion: how many cycles are you using? perhaps you could avoid to reach the plateau of the curve (maybe 20 or 25 cycles), stop the running and to run the samples in an agarose gel because it could be helpful to see the differences within the standard curves samples and your experimental samples.

I hope to have been helpfully although It may be is difficult without more information.

Good luck!! . I am pretty sure that you will solve the problem soon.

Best Regards,

Javier

ChIP efficiency cannot be expected to be the same in every experiment. Therefore, it is important that you normalize EVERY ChIP experiment against "input" material (an aliquot of extract prior to ChIP) to computing a % of input metric.

Are you using sonicated or MNase-treated material? Do you know how much of your variation is due to noise in the technical replicates vs. changes across the biological replicates? Which shows greater variability, technical or biological replicates?

A wide variability within technical replicates calls the accuracy of the experiment into question. These should be very similar, as this is basically a measure of pipetting accuracy.

Once you confirm there's little variability within your technical replicates, you should take the technical mean as one value for the biological replicate. For instance, if you had 3 biological replicates, each one assayed in technical triplicate, your sample size would be 3, not 9. Hope this helps.

It's a problem called pseudoreplication - when you have multiple data points for the same trait taken from the same biological replicate (at the same time point). Although technical reps have an important role telling you how much variation arises from lab analysis, its important that you distinguish the two types in your statistical analysis (if technical variation is important).

I agree with Christopher, if the variability in your technical replicates is not important, then use the average of them, so that you have an average "technical rep" value for each of the 3 biological replicates. If you're not sure whether or not the variation between technical replicates is important, you could analyse it with an ANOVA. The ANOVA would have your 3 biological replicates, experimental factors (histone modification?), and also "technical replicate" as a factor. Hopefully technical replicate will not be a significant factor! If it is, then I another way could be a "nested" ANOVA, where you specify that each biological replicate has 3 technical replicates (this would recognise which 3 technical replicates came from each biological replicate. I hope that helps and makes sense.

One may compare the means of treatments of technical replicates of biological replicates using an array of statistical tests - ANOVA, Tukey, etc.!

We have a paper on analyzing qPCR data for gene expression with technical and biological replication using linear mixed models. Properly accounting for different levels of replication is essential for obtaining sound results.

The principles apply to any type of data (but of course you should follow proper data extraction/normalization protocols). We have a tool to do the analysis in SAS and the paper is cited in its webpage. Here:

https://www.msu.edu/~steibelj/JP_files/QPCR.html