I am investigating changes in gene expression in cells after light irradiation at multiple wavelengths using ddCt method. I would like to compare the control and irradiated areas and describe the increase or decrease in gene expression. Usually the Dunnett method is used for comparison, but I believe the ddCt values are not normally distributed.
What is the best multiple comparison method?
Hi Yuichi
instead of making RTq-PCR on a single gene to study differential expression, why could you not use arrays or more recently 3'RNAseq to see all changing genes in your samples?
all the best
fred
Parametric tests can be robust to certain levels of "nonnormal distributions" under certain conditions. Switching to non-parametric testing isn't necessarily a better way of analysis in that scenario. What does your ANOVA output look like, specifically QQ, Homoscedasticity, and residual plots look?
https://blog.minitab.com/en/adventures-in-statistics-2/choosing-between-a-nonparametric-test-and-a-parametric-test
https://statisticsbyjim.com/hypothesis-testing/nonparametric-parametric-tests/
Can Justin Kiessling ANOVA requires the assumption of a normal distribution, but can the ddCt values be considered normally distributed?
Frederic Lepretre Based on the results of the studies I have conducted so far, I thought that genes related to cell death were regulated in their expression levels. Therefore, instead of performing an exhaustive analysis, I performed qPCR to examine specific gene expression levels.
Hi
in a last paper I saw that 44 cell death genes were identified (based on array results) and I guess it could be more since it depends on lot of things, that's why I proposed this way.
https://www.nature.com/articles/s41419-024-06632-7
all the best
fred
1) I see that you used delta delta Ct, did you do any follow-up transformation (log2, etc)?
2) Have a look at the top reply by Bruce - https://www.researchgate.net/post/I_am_planing_to_do_ANOVA_test_on_my_data_but_the_data_are_not_normally_distributed_even_after_transformation_eg_log_Box-Cox_What_should_I_do/1 & https://www.researchgate.net/post/Can_the_use_of_a_parametric_test_ANOVA_after_a_failed_normality_test_be_justified
3) Without knowing the specifics of your case, it is hard to make further comments on how unsuitable an ANOVA-based method would be in your case. Strong violations could make some features such as linear regression analysis & descriptive values such as 95% CIs unsuitable way to describe that data if that's the way you wanted to describe your results/data, though with some modification it may be possible (https://online.stat.psu.edu/stat415/lesson/2/2.6).
4) Have you replicated your experiment, would be curious to see P-values of the original and replication experiment P-values. If you strong effect sizes (delta delta Cts/log fold change) with low P-values that are reproducible and make sense in your biological context, then I don't think small departures from normality are going to be raising alarm bells.
How did control genes (housekeeping) behave in different cell samples? How many did you try? Your ddCt might depend on this too much.
Frederic Lepretre Thank you so much sir. Actually, I use Drosophila cell line, not human. I previously read an article that is researched expressions of 28 genes involved in apoptosis using qPCR, so I tried qPCR.
Article Apoptotic Activity and Gene Responses in Drosophila melanoga...
However, as you said, more genes are involved in apoptosis in Drosophila as well as human. If there is no difference in my qPCR result, I'll try array.
- Badges
- Science topic
- Similar topics
- Filing