I worked on cancer cells. Added cytotoxic drug at three different concentration. Each concentration added in six replicates. I incubated the cells after adding PrestoBlue. Then read the flouscence of the plate. Shall I use ANOVA then post hoc as all groups related to each other or shall I do student t test between each concentration and the control?
Actually, you should design a better experiment.
Why did you use only 3 concentrations but 6 replicates at each concentration?
If your aim is to estimate a dose-response relationship (or simply to find a dose at which the response is maximal/minimal) you should have used more different concentrations, and there would no need to replicate any of these doses. So you could have 18 values for 18 different concentrations spanning the range of interesting concentrations.
What you are asking for here is more kind of a "post mortem analysis"...
Hi,
Generally, multiple t-tests are discouraged as the probability of a significant result simply due to chance (alpha error or Type I error) accrues. This can be partly corrected with Bonferroni correction or ANOVA with posthoc pairwise comparisons.
But first, you need to be clear of the hypothesis you are testing. If you are trying to find the optimum concentration to control untreated cancer cells, perhaps testing a wider range of concentrations and establishing a dose-response curve would be more useful.
Hope it helps!
Dear Heba,
My recommendation follows more or less those ones of Jochen and Qin mentioned above with a few additional suggestions. The survival of cells (as percentage: f(S)) in dependence on the (presumably) cytostatic compound can be described by the following sigmoidal function:
f(S) = a0 * 1/{1 + e^(a1 - a2*[c(C)]}.
In this equation, a0, a1, and a2 are approximating coefficients, which can be determined by non-linear regression (available in almost all statistical software packages, which provide you also with the significance levels for each variable coefficient). c(C) is your concentration of the compound in question. See e.g. https://www.researchgate.net/publication/13686930_A_low_concentration_of_ethanol_reduces_the_chemiluminescence_of_human_granulocytes_and_monocytes_but_not_the_tumor_necrosis_factor_alpha_production_by_monocytes_after_endotoxin_stimulation
Your half-maximal lethal dose (key characteristics of the cytotoxicity) can be determined by setting the second ablation of the equation above to zero.
As you have three coefficients to determine, you need at least six different concentrations that cover the entire range of cell survival (from 90%) as your degrees of freedom must not drop below 3.
I personally would go for 3 repetitions (inter-day variance) with 7 concentrations.
ANOVA is not the right approach - there the grouping is categorical, meaning that you loose the information that you add always the same compound at different concentrations.
Please be aware of that the cytotoxicity does not correspond to the toxicity of your compound. Hence a compound killing the cells as such might inactivate key metabolic pathways and the results might be worth nothing if the same toxicity applies to all cells of the body. I'd rather go for inhibition of cell proliferation, applying either the equation mentioned above or using an ANCOVA if the necessary prerequisites (linearity after transformation) are met.
Good luck,
Alex
Article A Low Concentration of Ethanol Reduces the Chemiluminescence...
Hello! I hope someone will answer my query, which is related to this question. The experimental design is that I have only one control group: negative or untreated control. And then, I have 8 experimental groups, the 8 different concentrations of my test substance. When comparing the results of your control group versus the experimental group, which statistical tool is the best to use when solving its p-value? Thank you!
- Badges
- Science topic
- Similar topics
- Biological Science
- Molecular Cell Biology
- Culture Cells
- Cell Line