The accurate description of protein-ligand binding energies and configurations is an important problem in molecular biology with many applications in medicine and pharmacology. Molecular dynamics (MD) simulations provide the required accuracy but they are too slow for searching binding positions. Conversely, docking methods are much faster but have limited accuracy. An appropriate combination of the two methods could avoid the shortcomings associated with each, thus offering a better approach to the protein-ligand binding problem. Here we investigate the feasibility of such a combined docking-MD approach in a well-defined system: binding of organic cations to the gramicidin A channel. We use the AutoDock program to generate a set of protein-ligand binding configurations, which are then refined in MD simulations. For each system, we examine the binding configuration in detail and calculate the binding free energy by constructing the potential of mean force for the ligand. Our results show that AutoDock provides suitable initial configurations, which can be used profitably in MD simulations to obtain an accurate description of proteinligand binding with a reasonable computational effort.