Parameter estimation for a two-site sequential binding model. (A) Diagram of model in which macroscopic binding constants K₁ and K₂ quantify the affinities of the first and second binding steps, respectively. M_i refers to receptor with i-bound ligands, and L refers to ligand. (B) Simulated binding curves for two different parameter sets. Parameter set A is consistent with weak cooperativity between the binding sites (K₁ = 200 µM⁻¹ and K₂ = 600 µM⁻¹). Parameter set B is consistent with strong binding cooperativity (K₁ = 75 µM⁻¹ and K₂ = 1,500 µM⁻¹). Gaussian noise was added to the curve for parameter set B to mimic experimental variability. Although these parameter sets (and their mechanistic interpretations) are quite different, they produce nearly identical observables. (C) Log-error surface in K₁–K₂ parameter space with respect to the noiseless data curve in B. For visualization, log-error values are contoured at levels {−8.5, −8, −5, −4, −3, −2.5, −1.5, −1, 0, 1, 2}. Although this error surface is bounded, large ranges of parameter values produce very similar binding curves. (D) MCMC samples of joint posterior distribution of the parameters when constrained by the noisy curve in B.