Binding site prediction is a critical early step in a ligand discovery project. Computer-aided drug design (CADD) offers a variety of methods to identify putative ligand-binding sites on proteins and RNA. CADD workflows, which are complementary to experimental techniques, serve to narrow down a large problem in an efficient manner; for example, identify all possible binding sites and make predictions that can be validated using slower but more accurate techniques later in the pipeline such as more robust validation of binding sites appropriate for drug-like molecules. In this chapter, we first introduce the problem and the standard structure-based docking methods and then discuss machine learning methods including increasingly powerful deep learning tools. Finally, we discuss the physics-based molecular dynamics and cosolute methods, with special attention given to the Site Identification by Ligand Competitive Saturation (SILCS) technology. We discuss the advantages of the physics-based approach taken by SILCS and related methods.
Nordquist et al. (Thu,) studied this question.