I have a dataset containing IC50 data for each of my cell lines and up to 80 drugs. Basically for each drug I want to be able to define an IC50 threshold where if a cell line has an IC50 above that threshold it is resistant, and if it is below that threshold it is sensitive.

I'm trying to think of the best way of doing this: I was thinking of taking the median for each drug as the threshold but I realize that's quite crude. Are there any more robust statistical techniques I could use to define such a threshold for each drug?



IC50 has an exact definition, that is half-maximal inhibitory concentration. To get this you need to make a series of dilutions of each of your drugs and test your cell lines in each concentration and assess how they grow (eg measure Optical density, colony size, cell count or whatever applies to your cells and setup). Then you can set up a inhibition curve that is a function of the concentration. This most likely looks like a sigmoid curve and the inflection point gives you the IC50 value.

An example image form the wikipedia page of IC50 (recommended to read): enter image description here

So in your case, you already have IC50 values (for wild types I assume), and you need to assess your (probably mutant ?) cell lines to see if they are sensitive or resistant. Standard lab practice recommends using wild-type controls and at least 3 parallel controls. You can compare the average growth of WTs to your cell line at the IC50 concentration you already got in your dataset, and calculate standard deviation, and 95% confidence interval. With some simplification if the 95%CI of your cell line is greater than the wt 95%CI it is most likely that your cell line is resistant to that drug, on the other hand if its lower then that cell line is sensitive to the drug in question.


You might want to see the distribution of IC50s. If the majority do not obtain resistance, the histogram may show a normal distribution. Then you would know where less than 5% or 1% of population distribute. If you have a few cell lines know as resistant against the chemical, you could use them to decide threshold.


Each drug has different effective dosage. You can not draw the same threshold for different drugs. If you have lots of cell lines, you could apply the strategy I mentioned above for a single drug.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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