The evaluation of the inhibitory activities of drugs on multiple cardiac ion channels is required for the accurate assessment of proarrhythmic risks. Moreover, the in silico prediction of such inhibitory activities of drugs on cardiac channels can improve the efficiency of the drug-development process. Here, we performed molecular docking simulations to predict the complex structures of 25 reference drugs that were proposed by the Comprehensive in vitro Proarrhythmia Assay consortium using two cardiac ion channels, the human ether-a-go-go-related gene (hERG) potassium channel and human Na_V1.5 (hNa_V1.5) sodium channel, with experimentally available structures. The absolute binding free energy (ΔG_bind) values of the predicted structures were calculated by a molecular dynamics-based method and compared with the experimental half-maximal inhibitory concentration (IC₅₀) data. Furthermore, the regression analysis between the calculated values and negative of the common logarithm of the experimental IC₅₀ values (pIC₅₀) revealed that the calculated values of four and ten drugs deviated significantly from the regression lines of the hERG and hNa_V1.5 channels, respectively. We reconsidered the docking poses and protonation states of the drugs based on the experimental data and recalculated their ΔG_bind values. Finally, the calculated ΔG_bind values of 24 and 19 drugs correlated with their experimental pIC₅₀ values (coefficients of determination=0.791 and 0.613 for the hERG and hNa_V1.5 channels, respectively). Thus, the regression analysis between the calculated ΔG_bind and experimental IC₅₀ data ensured the realization of an increased number of reliable complex structures.