I have used different methods including graphpad prism, excel and various other online tools but i get different values each time. I want to know how IC50 can be determined from % inhibition and concentrations.
I find GraphPad Prism to be the best tool when calculating ic50.
Using the XY analysis ( points and connecting line graph), you should plot log concentrations on the X-axis vs % viability (determined previously on excel = average sample value/ average control *100) on the Y-axis. Thereafter, analyse using the XY tool: dose-response inhibition (start with log inhibitor vs normalised response); you can also try the other options; the ic50 should correlate to the extrapolation on the graph- that is how you determine the accuracy of the result.
Different IC50s will result from the same data if the various nonlinear regression analyses are using different equations, or if in some cases logarithms of the concentrations are used instead of the original inhibitor concentrations and the number of decimal places is truncated in the logarithms. Another difference between methods is whether you use % inhibition data or the raw measurements.
The simplest equation for nonlinear regression to obtain an IC50 is the 2-parameter Hill equation, which assumes that the % inhibition is zero when there is no inhibitor and the maximal inhibition is 100%. The two parameters to fit are the IC50 and the Hill coefficient (n). % inhibition = 100[I]n/(IC50n+[I]n) where [I] is the inhibitor concentration.
If there is a reason to believe that the maximal % inhibition (MAX) is not 100%, then a 3rd parameter is needed, replacing 100 with MAX. Some versions of the Hill equation may have a 4th parameter, a constant term, which is a nonzero % inhibition at zero inhibitor concentration. This may improve the fit to the data, but it doesn't correspond to reality. It most likely reflects a problem with the positive and/or negative controls used to calculate % inhibition.
To calculate % inhibition, you need two controls: a positive control (no inhibitor) and a negative control (complete inhibition).
If you tell me what the expected relationship between your variables is, I can check if the appropriate model is already in my software. If not, I can add it for you (no charge, but I will need your data to check the model).