Determining the effect of a compound on I (Kr) is a standard screen for drug safety. Often the effect is described using a single IC(50) value, which is unable to capture complex effects of a drug. Using verapamil as an example, we present a method for using recordings from native myocytes at several drug doses along with qualitative features of I (Kr) from published studies of HERG current to estimate parameters in a mathematical model of the drug effect on I (Kr). I (Kr) was recorded from canine left ventricular myocytes using ruptured patch techniques. A voltage command protocol was used to record tail currents at voltages from -70 to -20 mV, following activating pulses over a wide range of voltages and pulse durations. Model equations were taken from a published I (Kr) Markov model and the drug was modeled as binding to the open state. Parameters were estimated using a combined global and local optimization algorithm based on collected data with two additional constraints on I (Kr) I-V relation and I (Kr) inactivation. The method produced models that quantitatively reproduce both the control I (Kr) kinetics and dose dependent changes in the current. In addition, the model exhibited use and rate dependence. The results suggest that: (1) the technique proposed here has the practical potential to develop data-driven models that quantitatively reproduce channel behavior in native myocytes; (2) the method can capture important drug effects that cannot be reproduced by the IC(50) method. Although the method was developed for I (Kr), the same strategy can be applied to other ion channels, once appropriate channel-specific voltage protocols and qualitative features are identified.