I have three biological repeats of my experiment but the variance across the experiments is very high (although the trend among my treatment groups in each experiment is consistent). Does anyone know how to normalize across multiple MitoStressTest experiments? Any helpful references on data normalization for in vitro studies would be appreciated!
Thank you!
- Sindhu
Could you run your experiment again with three replicates? Do you get the same result/conclusions? Do not combine the data sets to get 6 replicates, because at this point you are asking if another scientist runs your experiment with three replicates what is the probability that they will arrive at the same conclusions.
In general, three replicates per treatment is not sufficient. Power is low, results are not reproducible, and predictive value is nil. These are all symptoms of the root cause that one sample of three replicates will do a poor job of estimating population values even if you are correct in assuming a normal distribution.
@timothy - just to clarify: i have 3 treatment groups (control, treated, knockdown + treated) and each experiment (Mito Stress Test XFe 24) had 2-4 technical replicates for each treatment condition. so across all three experiments (3 biological replicates), each treatment group had 6-12 measurements. so i'm wondering how to look at statistical significance across all three experiments (i.e. what is the best way to normalize across experiments, which statistical test to use etc).
in case it helps anyone else - so far what i have done is to normalize all measurements in each experiment to the average basal respiration of my control samples for that experiment. next, i averaged the three measurements after each drug (i.e. avg basal, avg oligo, avg fccp, and avg rot/aa). and then i calculated all other seahorse/mito stress test parameters (atp production, spare capacity, proton leak, non-mito ocr etc) using the avg values. after repeating these calculations for each seahorse experiment, i used the students t-test to compare treated vs. control and knockdown/treated vs. treated across the 3 experiments. this gives me an idea of how atp production is changing with respect to the basal respiration rate.
There are several ways to normalize seahorse data, like DNA content, protein levels, cell numbers, etc. You can choose whatever is feasible and reliable in your case.
I think, if you standardize data with basal respiration rate, then you may loose some absolute information. Sometimes, whole respiration curve shifts up or down due to treatment. I don't know what is your case, but I would recommend to use both absolute and relative data.
A small note on t-test... I think it would be better to use ANOVA than t-test in your case.
Excuse me, if I have missed something here. Best luck!
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis