Dynamic Docking, Path Analysis and Free Energy Computation in Protein-Ligand Binding

Comprehending how drugs interact with biological macromolecules to form a complex with consequent biological response is particularly relevant in drug design to guide a rational design of new active compounds. The establishment and the duration of the protein-ligand binding complex is principally determined by thermodynamics and kinetics of the dynamical process of molecular recognition. Thus, an accurate characterization of the free-energy governing the formation of the protein-ligand complex is of fundamental importance to deeply understand each contribution to the establishment of the molecular complex. Experimental biophysical techniques proved to be efficient in characterizing both thermodynamics and kinetics of protein-ligand binding. However, a detailed description of the whole binding process on a mechanistic level is not possible since only a quantitative estimation is allowed. Conversely, from the computational point of view, plain molecular dynamics, which has been increasingly considered as the method of choice to investigate the entire dynamic process upon complex formation and to predict the associated thermodynamic and kinetic observables, cannot be applied in a routinely drug discovery pipeline because of the high computational cost. In this context, this PhD thesis wants to address specific aspects of the protein-ligand binding process. In particular, it will deal with dynamic docking, thermodynamics and kinetics of protein-ligand binding by devising respectively three different computational protocols. We developed a dynamic docking protocol based on potential-scaled (sMD) simulations, in which the protein and the ligand are let completely flexible in order to predict the protein-ligand binding pose within a reasonable computational time. Then, we investigated the applicability of sMD in describing the kinetic behavior of a series of drug-like molecules and we devised a fully automated method to analyze the unbinding trajectories. Finally, we develop a semi-automated protocol based on path collective variables combined with well-tempered metadynamics to estimate free-energies along a binding path.