We are currently doing an screening where supernatans from a bacteria genus are tested againts a pathogenic bacteria cell culture in a 96-well plate where OD is measured each 2 hours.

Each supernatant from a strain has 3  replicates  and the control phatogenic bacteria cell culture has about 11 replicates inside the plate.

Once each experiment is finished cell growth curves are modelated and 3 growth parameters are determined (mu, lambda and max absorbance) from each supernatant tested and from the control. Then a Mann-withney U test is performed for each growth parameter from each strain againts corresponding control parameter to check if there's some kind of possible inhibition.

The strains which has some inhibition efect againts this pathogenic bacteria will be tested again in further experiments to confirm the possible positive.

I  have a doubt about if performing statistical test to compare supernatants from each plate againts its corresponding control is correct considering that the possible candidates will be again tested in further experiments to confirm its inhibition. Since there's loads of strains to test, repeating each single experiment more than 3 times would take too long.

What it is your oipinion? Is it correct to do statistics for single plate (technical replicates) to gather possible positive for a later confirmation (biological replicates)?

Are biological replicates strictly needed in this case?


1 Answer 1


Given that this is a screening assay where any apparent positive results will be tested in new (presumably larger) experiment, there may be no need to perform a statistical test. Any test that you do use is likely to have low power given the small number replicates and so you would quite likely erroneously discard many of the candidate strains.

Given that you will have more data about the control than the test wells, consider calling a putative positive any supernatant that gives an average maximum or lambda that is more than a standard deviation or so different from the controls. If that gives too many putative positives then increase the threshold accordingly.

It is important to have control wells on every plate, and ideally you would have the well locations randomised. "Biological replicates" are going to be essential in the follow-up experiments if you wish to make inferences that extrapolate beyond the single plate being tested, but for the preliminary screen they may be an inefficient use of resources.

(Please run a spell checker!)


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.