Predicting binding energetics from structure: looking beyond DeltaG degrees.

Abstract

Structure-based design of pharmaceuticals requires the ability to predict ligand affinity based on knowledge of structure. The primary term of interest is the binding affinity constant, K, or the free energy of binding, DeltaG degrees. It is common to attempt to predict DeltaG degrees based on empirically derived terms which represent common contributions such as the hydrophobic effect, hydrogen bonding, and conformational entropy. Although these approaches have met with some success, when they fail it is difficult to know which parameter(s) need refinement. Confidence in these approaches is also limited by the fact that DeltaG degrees typically is made up of compensating enthalpic and entropic terms, DeltaH degrees and DeltaS degrees, so that accurate prediction of a DeltaG degrees value may be fortuitous and may not indicate a reasonable understanding of the underlying relationship between structure and affinity. This is further complicated by the fact that both DeltaH degrees and DeltaS degrees are strongly temperature dependent through the heat capacity change, DeltaCp. In order to avoid these difficulties, we attempt to use structural data to predict DeltaH degrees, DeltaS degrees, and DeltaCp from which DeltaG degrees can be calculated as a function of temperature. The predictions are then compared to experimentally determined values. These calculations have been applied to several systems by ourselves and others. Systems include the binding of angiotensin II to an antibody, the dimerization of interleukin-8, and the binding of inhibitors to aspartic and serine proteases. Overall the calculations are very successful, and suggest that our understanding of the contributions of the hydrophobic effect, hydrogen bonding, and conformational entropy are quite good. Several of these systems show a strong dependence of the binding energetics on pH, indicative of changes in proton affinity of ionizable groups upon binding. It is critical to account for these protonation contributions to the binding energetics in order to assess the reliability of any computational prediction of energetics from structure. Methods have been developed for determining the energetics of proton binding using isothermal titration calorimetry. The availability of these methods provides a means of understanding how protein structure can modify the pKa's of ionizable groups. This information will further add to our understanding of structural energetic relationships and our ability accurately to predict binding affinities.