Molecular recognition and drug-lead identification: what can molecular simulations tell us?

Molecular recognition and ligand binding involving proteins underlie the most important life processes within the cell, such as substrate transport, catalysis, signal transmission, receptor trafficking, gene regulation, switching on and off of biochemical pathways. Despite recent successes in predicting the structures of many protein-substrate complexes, the dynamic aspects of binding have been largely neglected by computational/theoretical investigations. Recently, several groups have started tackling these problems with the use of experimental and simulation methods and developed models describing the variation of protein dynamics upon complex formation, shedding light on how substrate or inhibitor binding can alter protein flexibility and function. The study of ligand-induced dynamic variations has also been exploited to review the concept of allosteric changes, in the absence of major conformational changes. In this context, the study of the influence of protein motions on signal transduction and on catalytic activities has been used to develop pharmacophore models based on ensembles of protein conformations. These models, taking flexibility explicitly into account, are able to distinguish active inhibitors versus nonactive drug-like compounds, to define new molecular motifs and to preferentially identify specific ligands for a certain protein target. The application of these methods holds great promise in advancing structure-based drug discovery and medicinal chemistry in general, opening up the possibility to explore broader chemical spaces than is normally done in an efficient way. In this review, examples illustrating the extent to which simulations can be used to understand these phenomena will be presented along with examples of methodological developments to increase physical understanding of the processes and improve the possibility to rationally design new molecules.