Input:
- RT-qPCR dataset row CT values interpolate corrected
- 50 target genes
- 10 reference genes
- 3 different treatments with 3 replicates each
- 1 control replicate for each treatment
Objective:
- Identify differentially expressed genes between each treatment replicate and the corresponding control.
Already done:
- Calculated Delta Delta CT values and fold changes
Question:
- Some of the treatments are normally distributed, most of them are not.
Which statistical test would fit the data best in this case?
- Does it make sense to calculate the p-value for each gene in each treatment replicate?
- How should I apply the statistical tests to calculate the p-values considering all values for the same replicate?
- I was thinking about calculating the FDR corrected p-values afterwards. How would I do it in this case or is there a suitable alternative to this method?
Do you have biological replicates that are in triplicates? It seems like you have n=1 of control, ideally you would want at least 3. If you are expecting high variability, ideally you'd have more than 3 biological replicates i.e. gene expression in in vivo cancer cells may not be so homogeneous as in vitro cell lines. Since you have 3 groups, you can do ANOVA with Tukeys (normally distributed) or Kruskal-Wallis (not normally distributed) for FDR correction.
Hi,
I agree with Can Kiessling , you need biological replicates. You cannot get p-values from experimental groups which contain just one animal (biological replicate). Also, two technical replicates should be done for each biological replicate.
I was thinking how you evaluated the normality, which test you use? Shapiro-Wilk, D'agostino?
Your "n" is very small and i do not think you can be sure about data normality, use Kruskal-Wallis with Dunns post test or PERMANOVA (distribution free test). Softwares: GRAPHPAD PRISM8, SAS, STATISTICA, R.
- Does it make sense to calculate the p-value for each gene in each treatment replicate? P-values are obtained for each comparison between groups you make (Control versus Treatment 1, Control versus Treatment 2...). Inside the same treatment group does not make sense.
- How should I apply the statistical tests to calculate the p-values considering all values for the same replicate? Replicates are experimental units were you apply you treatment. If a inject something in a rat, each rat i inkected is a biological replicate.Imagine this experimental groups: Control (3 rats), DRUG A (3 Rats), DRUG B (3 Rats) DRUG C (3 Rats). Here, we have 4 experimental groups (Control, DRUG A, DRUG B and DRUG C with three biological replicates each.
Best,
Jacó
Greetings,
I agree with Jaco and Can. Once your experimental design is corrected. Your can always use ANOVA or Kruskal-Wallis to get basic statistics. However, I'd recommend you to use standard protocols to determine the suitability of your candidate HK genes, like geNorm, NormFinder and BestKeeper. You could even compare the results with Livaak method, or any other available at published papers, and use Pearson's correlation to correctly assess the outcome of your experiment.
Good luck
- Badges
- Science method