Three-dimensional structures of both the open- and closed-channel states of alpha 4 beta 2 receptor have been modeled and used to study their binding with representative agonists and antagonists. The obtained binding structures and free energies consistently reveal that antagonists bind more favorably with the closed-channel state and agonists bind more favorably with the open-channel state. The computational insights have led us to propose a computational strategy and protocol predicting whether a receptor ligand is an agonist or antagonist. Using the computational protocol, one only needs to calculate the relative binding free energies for a ligand binding with the open- and closed-channel structures. The ligand is predicted to be an agonist if the binding free energy calculated for the ligand binding with the open-channel state is significantly lower than that for its binding with the closed-channel state. If the binding free energy of a ligand with the open-channel state is higher than that with the closed-channel, the ligand is predicted to be an antagonist. The binding free energies calculated for all of the ligands binding with their most favorable channel states of the receptor are all close to the corresponding experimentally derived binding free energies. The new computational insights obtained and novel computational strategy and protocol proposed in this study are expected to be valuable in structure-based rational design of novel agonists/antagonists of nAChRs as therapeutic agents.