The energy landscape approach has contributed to recent progress in understanding the complexity and simplicity of ligand-macromolecule interactions. Significant advances in computational structure prediction of ligand-protein complexes have been made using approaches that include the effects of protein flexibility and incorporate a hierarchy of energy functions. The results suggest that the complexity of structure prediction in molecular recognition may be determined by low-resolution properties of the underlying binding energy landscapes and by the nature of the energy funnels near the native structures of the complexes.