Structural Parameterization of the Binding Enthalpy of Small Ligands Irene Luque and Ernesto Freire * Department of Biology, The Johns Hopkins University, Baltimore, Maryland ABSTRACT A major goal in ligand and drug design is the optimization of the binding affinity of selected lead molecules. However, the binding affin- ity is defined by the free energy of binding, which, in turn, is determined by the enthalpy and entropy changes. Because the binding enthalpy is the term that predominantly reflects the strength of the inter- actions of the ligand with its target relative to those with the solvent, it is desirable to develop ways of predicting enthalpy changes from structural consid- erations. The application of structure/enthalpy cor- relations derived from protein stability data has yielded inconsistent results when applied to small ligands of pharmaceutical interest (MW < 800). Here we present a first attempt at an empirical parameter- ization of the binding enthalpy for small ligands in terms of structural information. We find that at least three terms need to be considered: (1) the intrinsic enthalpy change that reflects the nature of the interactions between ligand, target, and solvent; (2) the enthalpy associated with any possible conforma- tional change in the protein or ligand upon binding; and, (3) the enthalpy associated with protonation/ deprotonation events, if present. As in the case of protein stability, the intrinsic binding enthalpy scales with changes in solvent accessible surface areas. However, an accurate estimation of the intrin- sic binding enthalpy requires explicit consideration of long-lived water molecules at the binding inter- face. The best statistical structure/enthalpy correla- tion is obtained when buried water molecules within 5–7 Å of the ligand are included in the calculations. For all seven protein systems considered (HIV-1 protease, dihydrodipicolinate reductase, Rnase T1, streptavidin, pp60c-Src SH2 domain, Hsp90 molecu- lar chaperone, and bovine -trypsin) the binding enthalpy of 25 small molecular weight peptide and nonpeptide ligands can be accounted for with a standard error of 0.3 kcal  mol 1 . Proteins 2002;49: 181–190. © 2002 Wiley-Liss, Inc. Key words: binding energetics; binding thermody- namics; isothermal titration calorim- etry; structure-based drug design; HIV-1 protease INTRODUCTION The completion of the Human Genome Project as well as the genomes of several pathogens has generated new challenges in ligand and drug design. The number of targets for drug development is expected to increase dramatically during the next few years. For each new target, lead compounds will need to be identified and optimized in order to achieve the required binding affinity, specificity, selectivity, bioavailability, and toxicological properties. This new reality accentuates the need for improved design paradigms and efficient ways of predict- ing binding parameters from structure. One approach to the prediction of binding parameters that does not require extensive computations is the development of accurate empirical parameterizations based on structure/thermody- namic correlations. Previously, such approach has been successfully applied to protein stability, 1,2,7,8 and its pre- dictive ability has been demonstrated in several laborato- ries. 1–4 Currently, the binding affinity is used as the main selection criteria in high-throughput screening and compu- tational analysis. Nonetheless, the binding affinity is defined by the Gibbs energy of binding (G), which, in turn, is determined by the enthalpy (H) and the entropy (S) changes (G =H - TS). In principle, many combinations of H and S can give rise to the same G value and, therefore, to the same binding affinity. Because the magnitudes of the enthalpy and entropy changes reflect different underlying interactions, ligands that have been enthalpically or entropically optimized exhibit differ- ent responses to changes in the target or the environment, even if they have the same affinity under the initial set of conditions. 5,6 For those reasons, additional selection and optimization criteria that includes enthalpic and entropic contributions have been proposed. 5,6 Here we describe a first attempt at an empirical struc- tural parameterization of the binding enthalpy for small ligands (MW  800). As in the case of protein stability, the basic requirement for a statistical parameterization is the availability of a set of protein complexes for which high- resolution structures and binding thermodynamic data are available. Because the conformational change in the Grant sponsor: National Institutes of Health; Grant number: GM 57144. Grant sponsor: National Science Foundation; Grant number: MCB9816661. Grant sponsor: postdoctoral fellowship from the Funda- cio ´n Ramo ´ n Areces, Madrid (Spain) to I. L. *Correspondence to: Ernesto Freire, Department of Biology, The Johns Hopkins University, Baltimore, MD 21218. E-mail: ef@jhu.edu Received 27 March 2002; Revised 16 May 2002; Accepted23 May 2002 Published online 00 Month 2002 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/prot.10206 PROTEINS: Structure, Function, and Genetics 49:181–190 (2002) © 2002 WILEY-LISS, INC.