Dose–response and IC50
Plan a concentration–response experiment: the dilution series, the replicates and the noise. See how precisely it estimates the IC50, and whether your concentrations bracket it at all.
More options
Fix the top and bottom at 100% and 0%?
Within two-fold of the truth
98%
Of 400 simulated experiments.
Typical estimate
1.01
95% of estimates between 0.57 and 1.52.
Fits that failed
0%
Lowest concentration
0.046
Wells per curve
24
One simulated experiment
Estimated IC50 across experiments (log10)
Teal bars are within two-fold of the true IC50.
For your protocol
Each concentration–response experiment will test 8 concentrations in a 3-fold dilution series from 100 down to 0.046, in 3 replicates, and the IC50 will be estimated by a four-parameter logistic fit. Simulation of 400 experiments, with a true IC50 of 1.00, a Hill slope of 1 and measurement noise of 8% of the response span, gave a median estimate of 1.01 (95% of estimates between 0.57 and 1.52); 98% of experiments estimated the IC50 within two-fold of the truth and 0% gave no usable fit.
How this simulation works
Responses follow a four-parameter logistic curve (bottom, top, IC50, Hill slope). Each simulated plate measures every concentration in replicate with normally distributed noise, and is fitted by least squares with the Levenberg–Marquardt algorithm on the logarithm of the IC50, with the asymptotes estimated or fixed at 0 and 100 for normalised data. A fit whose IC50 lands more than a hundred-fold outside the tested range counts as failed. 400 plates per setting. Real assays may also show plate effects, edge effects and outliers, which call for controls and robust fitting.