A Priori Prediction of Ligand Affinity by Energy Minimization

Abstract

Quantitative a priori prediction of the affinity of a ligand for its receptor is, in a sense, the holy grail of structure-based molecular design, ideally allowing one to optimize the number and kind of compounds designed for a program prior to any chemical synthesis. Historically, several approaches have been employed to calculate and/or predict binding affinity. The free energy of binding can be calculated directly via Free Energy Perturbation (FEP) calculations [1]. The relative binding energies for pairs of inhibitors are determined using a thermodynamic cycle in which the structure of one inhibitor is perturbed into the structure of another, both in the receptor site and in solvent. This approach has been reported to yield relative free energies that are accurate to ± 1 kcal/mol with respect to experiment. However, due to the amount of computer time required, these calculations are impractical for routine assessment of the binding affinity of proposed compounds. Other more efficient approaches have included: Comparative Molecular Field Analysis (CoMFA) [23], the Hypothetical Active Site Lattice (HASL) [4], HINT hydrophobicities [5], solvent accessibility [6] or solvent-induced interactions [7], atomatom contact preferences [8,9], a general mean field model [10,11] and scoring algorithms for database search or de novo design methods [12–16], as well as energy minimization methods. This chapter will focus on computational studies which employ energy minimization in an X-ray or modelled active site as the means to predict the affinity of a ligand for its receptor. These studies fall into two categories: (a) energy-component approaches – i.e. those that incorporate receptor-ligand energy values as one term in a sum of binding energy contributions or as part of a 3D QSAR; and (b) energy-only approaches – i.e. those that employ energy minimization methods alone to predict ligand affinity.