In my research I’m investigating 3 genotypes with 11 time points during a process. 8 gens where examinee and 18s served as a reference gene. The methodology of the qRT-PCR based on Michael W. Pfaffl (2001) (A new mathematical model for relative quantification in real-time RT-PCR) article. The data of the experiments analyzed with the software of the device (one step system) and the statistical analysis were conducted with DataAssist V3.01 software. Only a few points during the process are significant (p0.05). My question is: what’s the best software to analyze the data? This software should take into consideration the "delta delta ct" which is compared to the reference gene (18s).
Thanks
I suggest using the delta-delta-ct method ("ddct"). It requires that the amplification efficiencies of the target gene and the reference gene are identical. If they are not, you will anyway have a big pack of problems - whatever method you use.
You can model the effects of your predictors using a linear model on the dct values. That's all.
The bootstrap method is cumbersome and not powerful for small data sets. The argument to use the bootstrap method is that the fold-changes are not normal distributed. But the log-fold changes are. And ddct-values are log fold-changes. So dct values can be analyzed with standard methods (like linear models).
1. As with every statistical analysis you need to ask the right question and analyze accordingly. Which is the most important effect for you (genotype, time or gene)? Are effect dependent or not?
2. Fazels answer harbors a need in 3 factorial ANOVA and then post hoc multiple comparison tests which will eat away any signal you have. Worse, if you have results normalized to "1" the test might need to be non parametric. (SPSS or SAP or R)
3. I would advise against all that mess... You are welcome to contact, I'd be glad to help formalize the questions and design the analysis.
For statistical analysis of RT-qPCR we use a Bootstratio approach (I included the links to the publication and the web where you could do the calculations). I do not sure if it feets exactly your needs. You are welcome to contact if some advise in using it is needed.
http://rht.iconcologia.net/stats/br/
Article BootstRatio: A web-based statistical analysis of fold-change...
I found it very useful to use the free statistical software package fgStatistics:
https://sites.google.com/site/fgstatistics/down
It is relatively easy to use; you should read the Manual first, but it isn´t complicated.
Note that if you have many comparisons, you should carry a post test to avoid false positives. For example, at the significance level of p=0.05, you would expect to find only by chance one statistically relevant difference once in 20 comparisons. To avoid that problem, a False Discovery Rate should be done (Benjamini and Hochberg´s, FDR, I suggest):
Benjamini, Y. and Y. Hochberg, Controlling the False Discovery Rate: A Practical and
Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society. Series B (Methodological), 1995. 57(1): p. 289-300.
Good luck!
I suggest using the delta-delta-ct method ("ddct"). It requires that the amplification efficiencies of the target gene and the reference gene are identical. If they are not, you will anyway have a big pack of problems - whatever method you use.
You can model the effects of your predictors using a linear model on the dct values. That's all.
The bootstrap method is cumbersome and not powerful for small data sets. The argument to use the bootstrap method is that the fold-changes are not normal distributed. But the log-fold changes are. And ddct-values are log fold-changes. So dct values can be analyzed with standard methods (like linear models).
I normally use manual delta delta calculation in excel and it works better and you have the option of manupilating your data to suit your study. If you have a huge data set you can use excel macro to automate the calculation.
happy to answer if you have further questions.
Article Measuring the efficacy of anti-malarial drugs in vivo: Quant...
Thank you all... the methods are good but I'm looking for more methods...
Analysing qRT-PCR is not too difficult. There's a really good description of how to do it in this Nature Protocols paper (attached).
What is critical, however, is that the reference gene is appropriate to your genes of interest. If they have very different levels of expression the results may not be meaningful.
Finally, if you're wanting other software simply because some of your genes are p>0.05, then I'd worry. Don't use a package or software based on the results you want to achieve. If you've done the experiment correctly and used the correct statistical method then those gene are not significantly differentially expressed. Accept it and move on.
I know it's been almost two years after the last response in this topic, but I'm terribly confused at this point. From my really small experience, I have seen that ddCt data seems to be normally distributed, what seems not to occur with 2^-ddCt values.
Likewise, I have seen some authors saying that you should compute some statistics based on ddCt alone, and only then transform it to 2^-ddCt for reporting the results in a more clear vocabulary (fold change). "Since Ct is the observed value from experimental procedures, it should be the subject of statistical analysis. The practice of performing statistical analysis at ratio directly is not appropriate. The presentation of data needs to refer to the ΔΔCt and subsequently the ratio and confidence intervals derived from 2-ΔΔCt" (Yuan et al. 2006 - http://bmcbioinformatics.biomedcentral.com/articles/10.1186/1471-2105-7-85).
Even the original Livak article (2001) states that by saying the 95%CI and standard errors should be calculated using ddCt, and only then transformed. I have seen some previous posts from Dr. Jochen Wilhelm and it seems that he also agrees with that.
However, now I'm seeing this Nature Protocol from Schmittgen and Livak (2008), which I didn't know of, that explicitly says that "It should be emphasized
that statistical tests should not be run on the raw CT data and
s.d. should always be calculated after the 2DDCT , 2DCT or 2CT
transformation has been performed" (https://www.researchgate.net/file.PostFileLoader.html?id=5473089cd11b8bd30d8b46c5&assetKey=AS%3A273639499337734%401442252148686).
So can both be considered correct? This is a paradox to me.
As to what I have read are the authors of this paper not experts in statistics. The nature paper has several flaws regarding the analysis. And I agree that they contradict themselves once stating that "statistics should be done on dCt values" and once that "all statistics should be done on 2dCt values".
hi everyone...
what's what's difference between d-d-ct and Michael W. Pfaffl ?
- Badges
- Science topic