0
$\begingroup$

Take these for example:

https://doi.org/10.1167/iovs.16-19957 https://pubs.sciepub.com/ajcp/6/2/5/index.html

As far as I know, the formula for cell viability is % Viability = (Treatment OD - Blank OD) / (Control OD - Blank OD) where each OD is the average of three wells (technical replicates). Say I want to check the cytotoxic effects of increasing concentrations of a drug. The treatment wells in this case contain the drug, the vehicle, and cell medium for the rest of the volume. The controls in my experiment contain the corresponding amount of vehicle and cell medium but no drug. So to check 5 different concentrations of a drug, I have five controls containing the corresponding volume of the vehicle for each drug concentration, each of which is used to normalize the treatment groups (or alternatively, if I serially dilute the drug stock beforehand, one control group to normalize all treatment groups since all treatment groups will contain the same amount of vehicle). However, when I plot such an experimental design, I naturally get no bars for a zero concentration because the group does not exist. I may assume 100% for such a group for each trial and add them to the graph, but it would have 0 SD (and SE) since every value is 100%.

I tried to make a table for a design to make everything clearer: Table for the experimental design

Three plates with cells from different passages, so n = 3, each group is applied to three wells (technical replicates but their average is counted as one biological replicate). As you can see, you still get 100% from every trial for controls and their SD when plotted is still 0.

How then, do the researchers of these experiments get error bars for zero drug groups while at the same time normalizing their treatment groups to obtain a % cell viability? What sort of design do they use?

$\endgroup$
1
  • 1
    $\begingroup$ They're probably (incorrectly) assuming equal variance among groups. If you want to know what method they used read the methods section: what do they say? If they say nothing then they're not doing science. $\endgroup$
    – Bryan Krause
    Jan 18 at 14:33

1 Answer 1

0
$\begingroup$

In the shown data analysis your are taken average and standard deviation after normalizing. So evidently you will have mean = 100%, std = 0.

You can take average and standard deviation in the raw data (1.87, 1.808, 1.787, mean = 1.821, std = 0.0352). Then you will have a non zero std in the control group. Then you can normalized by the control average (mean = 100%, std = 1.93%).

$\endgroup$
1
  • 1
    $\begingroup$ I doubt they're doing this, they're almost certainly normalizing before doing any statistical analysis. I think they probably don't understand the analysis they performed and are just reporting the output of their statistical software blindly. $\endgroup$
    – Bryan Krause
    Jan 18 at 15:20

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .