The inactivation curves were done by the protocols described in Materials and methods and are fitted with a Boltzmann function 1/(1 + exp((V−V_1/2)/k)). For the sake of simplicity, we fixed the slope factor k to determine the shift of V_1/2 (the difference between V_1/2 values in the presence and absence of 100 μM carbamazepine). This simplification seems to be appropriate, as the slope of the curve does not greatly differ among these mutations either in the presence or absence of carbamazepine. V_1/2 shift of the wild-type channel is 6.1 ± 0.2 mV. The decrease in V_1/2 shift by mutation is then given as the difference between the shift in each mutant and 6.1 mV. The free energy change (ΔΔG) is directly derived from the product of the decrease in V_1/2 shift and the apparent equivalent gating charges (derived from RT/Fk or 25 mV/k). The interaction energy ΔΔG_inter is equal to the ΔΔG in the double mutant minus the summation of ΔΔG in each component single mutant. All of the ΔΔG_inter values are significantly smaller than zero, strongly suggesting the non-additive nature for W1716 and F1764 mutations to alter the drug binding affinity.