4.12 – Docking and Scoring

This chapter gives an introduction to small-molecule receptor docking and illustrates basic principles, challenges, potential pitfalls, approaches to improving the results, and successful applications. Small-molecule docking or structure-based virtual screening (VS) essentially consists of two steps: (1) first a small molecule is placed in the receptor, by exploring its external and internal degrees of freedom; (2) once one or more reasonable orientations have been found, the affinity of this molecule for the receptor is predicted, with the ultimate aim of comparing it to other molecules and differentiating good binders from bad binders. Some general considerations and the motivations for performing a VS campaign are outlined in Section 4.12.1 of this chapter. In Section 4.12.2 the innards of three docking programs are described in detail in order to give an idea of how docking works and what the algorithmic approaches are that can be taken. The algorithms selected are representative of a variety of approaches taken in this field, and cover methods such as Monte Carlo sampling, genetic algorithms, whole-ligand docking, and incremental buildup. There are many approaches to predicting the affinity of a molecule for its receptor, this process is generally referred to as scoring. In Section 4.12.3 scoring methods are analyzed in detail and various ways to improve scoring are described: tuning the scoring function to a certain target, designing a function that is firmly rooted in statistical mechanics, or combining several scoring functions. Although docking programs and scoring functions have been validated for general use, before embarking on a docking campaign against a specific target with a specific set of molecules, one should first assess which program and which scoring function (s) work best. Straightforward as this may seem, assessment of the quality of docking tools and scoring functions is fraught with unresolved issues. Section 4.12.4 contains considerations regarding evaluation and assessment and highlights some specific experiences of the authors. Section 4.12.5 describes the main challenges in structure-based VS. One key issue is that most docking programs do not adequately account for the fact that protein targets are flexible and, under physiological conditions, are immersed in water. Approaches to overcome this issue are described. Other improvements relate to accounting for nonspecific binders, increasing the efficiency of docking programs by introducing bias, and improving the quality of results by postprocessing. Several application examples are listed in the penultimate section (Section 4.12.6). Here the emphasis is on illustrating how structure-based VS is applied in modern drug discovery, often in combination with other in silico and experimental techniques. These examples range from high-throughput docking of a large number of compounds to applications where docking is one step in a cascade of in silico applications to winnow down a large set of molecules to a small set of potent binders. Finally we provide a glimpse of the future as we see it.