IC50 (half-maximal inhibitory concentration) is the concentration of a drug or compound that inhibits a biological process by 50%. It is calculated by fitting a 4-parameter logistic (4PL) sigmoidal curve to dose-response data. The IC50 is the inflection point of the sigmoid, where the response equals the midpoint between the Top (uninhibited) and Bottom (fully inhibited) plateaus. Lower IC50 values indicate more potent compounds.

IC50 and EC50 both describe the concentration at the sigmoid midpoint. IC50 (inhibitory concentration 50%) applies to inhibition curves where response decreases with increasing drug concentration, such as cell viability or enzyme activity assays. EC50 (effective concentration 50%) applies to activation or stimulation curves where response increases, such as agonist dose-response or cytotoxicity assays. Mathematically, both are the C parameter in the 4PL equation.

The Hill slope describes the steepness of the dose-response curve. A value of 1.0 indicates standard mass-action binding. Values greater than 1 indicate positive cooperativity or a steep transition. Values less than 1 indicate negative cooperativity or a shallow transition. Most single-target drugs have Hill slopes between 0.5 and 2.0. Hill slopes significantly outside this range may indicate multiple binding sites, assay artifacts, or compound aggregation.

A minimum of 6-8 concentrations spanning at least 2 log units above and below the expected IC50. The ideal experiment uses 8-10 half-log or full-log serial dilutions to capture both plateaus and the transition zone. At least 2-3 points should define each plateau, and 3-4 points should span the transition. Biological replicates (n=2-3) at each concentration improve confidence.

IC90 is the concentration that inhibits 90% of the biological response. It is calculated as IC90 = IC50 × 9^1/HillSlope. IC90 is used in antiviral drug development (predicts clinical efficacy better than IC50), antimicrobial susceptibility testing, and cancer combination therapy optimization. For a Hill slope of 1.0, IC90 is 9-fold higher than IC50.

A poor R-squared (below 0.95) usually indicates: incomplete curves missing one or both plateaus (add higher/lower concentrations); compound solubility limits causing precipitation; biphasic curves from two binding sites; excessive replicate variability; or too few data points. An R-squared above 0.95 is generally acceptable, and above 0.99 indicates an excellent fit.

IC50 is calculated by non-linear regression fit of a 4-parameter logistic (4PL) sigmoidal function to dose-response data on a log-concentration scale:

Y = Bottom + (Top − Bottom) / (1 + 10^{((logIC50 − X) × HillSlope)})

where X is log₁₀(concentration) and Y is the measured response. The four fitted parameters are Top, Bottom, logIC50 and HillSlope; IC50 is then recovered as 10^logIC50. Fitting on the log scale gives symmetric confidence intervals and is the form used by GraphPad Prism and most 4PL solvers. A well-fitted curve should have R-squared above 0.95, Hill slope between 0.5 and 2.0, and clearly defined Top and Bottom plateaus.

Typical IC50 values span roughly 6 orders of magnitude, from picomolar for the most potent biologics to millimolar for weak leads. Rough potency classes used in medicinal chemistry:

• Sub-nanomolar (< 1 nM): highly optimised clinical candidates and monoclonal antibodies.
• Low nanomolar (1–100 nM): typical drug-development leads and kinase inhibitors.
• Sub-micromolar (0.1–1 µM): early lead compounds.
• Low micromolar (1–10 µM): hit-to-lead compounds.
• Above 10 µM: weak hits that usually need further optimisation.

Cytotoxicity assays in cancer cell lines commonly report IC50 values between 10 nM and 10 µM. For most projects, a 10-fold IC50 improvement between hit and lead is a reasonable target.