Well, you need two, ideally parallel, dose-response curves. It's really best to use summary statistics such as IC50s for the selectivity index. You don't necessarily need to use the IC50 if you have plateaueffects or the like that make it difficult to estimate, an IC 10 might also work.

If you can't fit a proper simoidal curve you could also estimate from any linear parallel part of the two curves.

I agree with Dr Gremmels apart from the spelling of parallel.  Selectivity should only be calculated on a comparative, competitive basis.  However if you have a % cell inhibition for both compounds (say I1 and I2) at the same concentration, then the selectivity is simply the ratio of I1 to I2.

I stand corrected :-).

Is there any standardized range for selectivity index?

I have a question: Can IC50 values ​​greater than 100 be used as 100 to calculate the selectivity index?