Receptor flexibility in ligand docking

CAVASOTTO, CLAUDIO N.; ORRY, ANDREW J.; ABAGYAN, RUBEN A.

Handbook of Theoretical and Computational Nanotechnology

American Scientific Publishers

Año: 2006; p. 218 - 257

The Nobel Prize winners physiologist John Newport Langley (1852–1925) and immunologist Paul Ehrlich (1854–1915), working independently of one another, are widely accredited to be the first to introduce the concept of a receptor. Ever since these groundbreaking observations, scientists have been attempting to understand how molecules specifically interact with receptors. More recently, it has become clear that receptors are not static but undertake sometimes subtle but critical structural rearrangements on interaction with molecules within the environment of the cell. The formation of noncovalent complexes between molecules is crucial for the functioning of the biochemical machinery within a cell. Biological processes such as enzyme catalysis, molecular transport, and signal transduction rely on specific recognition between a receptor and the molecule to which it binds, otherwise known and referred to in this chapter as a ligand. In an attempt to understand the formation of these complexes, structural biologists have been using molecular atomic structure as a means of predicting and detecting the formation of ligand–receptor interactions. Determination of the structures of these interacting molecules on the microscopic level has led to the elucidation of the pharmacological effects of these complexes and resulted in the rational design of drugs. Recently, many worldwide structural proteomics initiatives and structure-based drug design methods have been having a significant effect on the development of therapeutics for a variety of diseases. Around 1000 new structures are being added to the Protein Data Bank (PDB) each year. These structures revealsome of the secrets of protein–ligand interactions, and by using this information with the development of new computational technology, we can design molecules that can bind to these proteins. Ligand docking into and virtualscreening (VS) of a drug target receptor is one method of identifying potentialdrug compounds based on the structure of the interacting molecules. The protein–ligand binding process is a complicated equilibrium between a solvated ligand and receptor in isolation and a solvated receptor–ligand complex. In the attempt to calculate the free energy of binding, many energy terms need to be considered in the equilibrium balance, including effects caused by breaking and forming hydrogen bonds, the hydrophobic effect, loss of translational, rotational and conformational entropy of ligand and receptor on binding, and so forth. The first attempt to predict molecular interactions was made in 1976. Since then, many new computational methods have been developed for protein docking, and although the reliability and accuracy of the methods have gotten better, there is still scope for improvement, particularly when a ligand induces some changes in the receptor on binding. Initially, docking methods were developed based on Fischer’s theory of a lock and key in which the three-dimensional (3D) structures of the ligand and receptor complement one another and fit favorably together. However, Koshland proposed a more accurate view of this process whereby the 3D structures of the ligand and receptor adapt to one another on binding, which is known as “induced fit.” The structure of the ligand modifies the receptor binding-pocket environment, which gives rise to many low-energy substates before binding. We review this theory and the modern explanation about the ligand binding process in Section 3.1. One of the main problems associated with computational protein docking is how to incorporate flexibility into the ligand and drug-binding pocket and how to appropriately dealwith “induced fit”. In general, a crystal structure is used as the basis of the structure-based design project; however, this only presents a rigid picture of the protein in the state in which it was crystallized, and not the many energetically favorable conformational states inherent in most protein receptors. The capability to predict many potentialconfor mationalstates of a receptor, for example, antagonist in its and agonist forms, enhances the accuracy of structure-based drug design. Many protein-ligand complexes including HIV protease, aldose reductase, and streptavidin support the “induced-fit” model. In Section 5, we will discuss algorithms that incorporate a degree of receptor flexibility into protein docking. Nowadays most docking algorithms can be used for virtual screening. However, incorporating flexibility into a virtual screening algorithm adds another level of complication to an already challenging problem. Conventional virtual screening processes uses a docking algorithm to screen any number of compounds, but generally in the range of from 200,000 to more than one million compounds, and then assess the docking scores. Different strategies can be used to reduce the poolof best scoring compounds (docking into a refined model, chemical clustering of the compounds, etc.) and the resulting compounds (usually numbering around 50–100) will be tested experimentally for their desired activity. Depending on the outcome of the experimental results, a new, focused virtual library can be developed from the highest-affinity compounds and rescreened, thereby refining the ligand scaffold and eventually increasing the experimental binding affinity. In Section 6, we will discuss algorithms that attempt to incorporate receptor flexibility into virtual screening process. In this chapter, we will discuss 1 current ligand docking and scoring methods; and why receptor and ligand flexibility is important for ligand docking; and current methods that incorporate flexibility into protein docking and virtual screening.