Hello Bahar Ksh ,

That's weird, your numbers look alright.

Maybe normalize 105 =1; 98 = 0.97...and so on?

Othwize there are R packages you can use for IC50 calculation and curve fitting, of course.

Hope that helps

Thomas

Hi Thomas Mühlenberg ,

Thanks for your reply.

Even after normalizing, because the numbers are not in order (like when it goes up at 80-->90 and then down again at 90-->72), GraphPad returns unexpected IC50 values.

Could you please introduce the R packages you mentioned?

Firstly, what are the drug concentrations at which you obtain the reported viability levels?

Secondly, have you measured viability in at least three replicates? If you calculate the mean ± SD, you may find that there is actually no statistically significant difference among the initial viability measurements.

So, if you leave out the outlier(s) (either 81, or 40, or both) Grphpad will deliver correct results?

The R package is Grmetrics

Hafner M, Niepel M, Chung M, Sorger PK: Growth rate inhibition metrics correct for confounders in measuring sensitivity to cancer drugs. Nat Methods 13:521-7, 2016

Hi Jakub Jagielski

The concentrations are: 0.1, 0.3, 1, 3, 10, 30, 100, 300

The viabilities are the average of three technical replicates. I will do biological replicates, but I can't get the IC50 for this assay. Maybe I should do 3 assays and then use their average fro IC50? But then that one number can not be used for statistical validation

Yes Thomas Mühlenberg , removing the outlier solves the issue. I wasn't sure if I should do that though.

Thanks for the packages!

For a publication you shouldn't. But as Jakub Jagielski suggested run at least triplicates (independent experiments) to even out outliers.

But if you want to get an idea or an intermetiate result, because your algorythm can't handle outliers, it's a way to move ahead. Especially in this specific case in which the general dose response relationship seems quite clear.

Let's assume your standard deviation (SD) for each average is 4.

So, this is how you calculate IC50:

• Plot viability (average ± SD) vs. concentration - interpolate concentration to a log10 scale.
• Fit a linear equation - you can use software such as OriginLab or even Excel. You will get: y = 86.49 (± 4.77) - 25.37 (± 3.50) * x, where y is viability, and x is log10 concentration (see below). In brackets are the standard deviations for each parameter.
• Calculate x (concentration) for y = 50 - you calculate at what concentration viability is 50%. In this example, you will get x = 1.433 ± 0.272.
• While the linear equation is fitted for log10 concentration, to get the actual concentration, you need to calculate 10^x, so 10^1.433 (± 0.272). You will get 27.1 ± 17.0. Take note of the high SD, therefore high uncertainty. However, notice also that we assumed SD = 4 for all samples, so there can be errors in calculations.
• If you take a look at the fitted graph, indeed viability is equal to 50 for a concentration of ~30, whatever units you use.

If you need help with more explanation, calculations and plots for actual values, feel free to contact me privately.

Thanks Thomas Mühlenberg , that makes sense.

Dear Jakub Jagielski , Thank you so much for the detailed explanation. That's very helpful. I'll repeat the test and then try this for all of the repeats.

Dear Bahar Ksh Can you check your data with the IC50 calculator given in the below link and see if you are able to get the answer?

https://mythreyaherbal.com/easy-ic50-calculator-for-herbal-drug-and-bioassay-studies/