Ligand-protein docking and rational drug design Terry P Lybrand University of Washington, Seattle, USA Over the past year there have been some interesting and significant advances in computer-based ligand-protein docking techniques and related rational drug-design tools, including flexible ligand docking and better estimation of binding free energies and solvation energies. As a result, the successful use of computational tools to help generate interesting new guide (lead) compounds for targeted receptors is becoming more commonplace. Current Opinion in Structural Biology 1995, 5:224-228 Introduction Computational tools Ligand-binding interactions (i.e. formation of a com- plex between two molecules) are central to numerous biological processes such as signal transduction, physi- ological regulation, gene transcription, and enzymatic reactions. Ligand-binding interactions encompass both macromolecular complexes (e.g. protein-protein and protein-DNA) and complexes of small molecules with macromolecules. As many proteins regulate key bio- logical functions via interactions with small molecules, these receptor proteins are often prime targets for thera- peutic agents. A detailed understanding of interactions between small molecules and proteins may therefore form the basis for a rational drug-design strategy. Ra- tional drug design is attractive as a drug development paradigm for two reasons: it offers some hope for re- duction of the enormous costs and time required in traditional random screening protocols for drug discov- ery, and may facilitate the development of more selective therapeutic agents with fewer undesirable side effects. Developments in molecular biology over the past 15 years make it possible to obtain experimentally use- ful quantities of numerous receptor proteins of po- tential therapeutic importance. Technical advances in X-ray crystallography, multidimensional NMtL spectr- oscopy, and other structural characterization methods have made it easier to obtain high-resolution struc- tural data for many important ligand-protein complexes. Still, there is at present no structural information for the vast majority of therapeutically interesting receptor pro- teins. In favorable cases, computational procedures such as homology model building may be used to generate approximate three-dimensional models for receptor pro- teins [1,2]. Development of various molecular modeling tools and ready availability of high-performance com- puting resources make possible detailed computational analyses and design projects for ligand-receptor com- plexes, provided suitable structural models are available for the receptors [3,4]. Over the past year, there have been some significant advances and improvements in computational tools used for ligand-protein docking and rational drug-design applications, and these devel- opments are the focus of this review. Interactive molecular graphics Numerous computational tools have evolved to inves- tigate ligand-receptor complexes and to develop new ligands [5]. Interactive molecular graphics methods have long been used to analyze receptor crystal structures and to manually dock candidate ligands in the binding site [6]. The interactive graphics approach is extremely la- bor intensive, primarily qualitative rather than quanti- tative in nature, and subject to personal biases. On the other hand, it takes advantage of the knowledge and in- tuition possessed by experienced structural biologists in a way that no strictly computational techniques have yet been able to do. As a result, interactive molecular graph- ics methods are used extensively and remain the principal tool for ligand design in many cases. Binding energy calculations Graphics model-building methods are often combined with energy calculations based on potential energy func- tions (see also this issue, Halgren, pp 205-210 and Sippl, pp 229-235 for a general discussion of current issues related to potential energy functions). Energy calcu- lations used routinely in rational drug-design applica- tions include energy minimization, molecular dynam- ics, Poisson-Boltzmann electrostatics, and free energy perturbation methods [7",8]. In principle, these meth- ods should provide quite detailed and definitive in- formation for rational drug-design projects, and there are many recent examples where these methods have been used to good effect [9-13,14°,15]. Inadequacies in potential energy functions and conformational sam- pling, however, restrict the power of these methods in modeling ligand-receptor complexes. These methods are also computationally quite expensive, which limits their practical utility in many cases. In some recent ligand-binding studies, attempts have been made to overcome certain limitations relating to potential functions and computational expense through the use of empirical free energy functions or free energy estimations [16,17]. These approaches often estimate the free energy of ligand-receptor interactions as a function 224 © Current Biology Ltd ISSN 0959-440X