Computational Approaches for Studying Drug Binding Kinetics

SEEKR2: Versatile Multiscale Milestoning Utilizing the OpenMM Molecular Dynamics Engine

We present SEEKR2 (simulation-enabled estimation of kinetic rates version 2)—the latest iteration in the family of SEEKR programs for using multiscale simulation methods to computationally estimate the kinetics and thermodynamics of molecular processes, in particular, ligand-receptor binding. SEEKR2 generates equivalent, or improved, results compared to the earlier versions of SEEKR but with significant increases in speed and capabilities. SEEKR2 has also been built with greater ease of usability and with extensible features to enable future expansions of the method. Now, in addition to supporting simulations using NAMD, calculations may be run with the fast and extensible OpenMM simulation engine. The Brownian dynamics portion of the calculation has also been upgraded to Browndye 2. Furthermore, this version of SEEKR supports hydrogen mass repartitioning, which significantly reduces computational cost, while showing little, if any, loss of accuracy in the predicted kinetics.

Machine Learning Analysis of τRAMD Trajectories to Decipher Molecular Determinants of Drug-Target Residence Times

Drug-target residence times can impact drug efficacy and safety, and are therefore increasingly being considered during lead optimization. For this purpose, computational methods to predict residence times, τ, for drug-like compounds and to derive structure-kinetic relationships are desirable. A challenge for approaches based on molecular dynamics (MD) simulation is the fact that drug residence times are typically orders of magnitude longer than computationally feasible simulation times. Therefore, enhanced sampling methods are required. We recently reported one such approach: the τRAMD procedure for estimating relative residence times by performing a large number of random acceleration MD (RAMD) simulations in which ligand dissociation occurs in times of about a nanosecond due to the application of an additional randomly oriented force to the ligand. The length of the RAMD simulations is used to deduce τ. The RAMD simulations also provide information on ligand egress pathways and dissociation mechanisms. Here, we describe a machine learning approach to systematically analyze protein-ligand binding contacts in the RAMD trajectories in order to derive regression models for estimating τ and to decipher the molecular features leading to longer τ values. We demonstrate that the regression models built on the protein-ligand interaction fingerprints of the dissociation trajectories result in robust estimates of τ for a set of 94 drug-like inhibitors of heat shock protein 90 (HSP90), even for the compounds for which the length of the RAMD trajectories does not provide a good estimation of τ. Thus, we find that machine learning helps to overcome inaccuracies in the modeling of protein-ligand complexes due to incomplete sampling or force field deficiencies. Moreover, the approach facilitates the identification of features important for residence time. In particular, we observed that interactions of the ligand with the sidechain of F138, which is located on the border between the ATP binding pocket and a hydrophobic transient sub-pocket, play a key role in slowing compound dissociation. We expect that the combination of the τRAMD simulation procedure with machine learning analysis will be generally applicable as an aid to target-based lead optimization.

Rate Constants and Mechanisms of Protein-Ligand Binding

Whereas protein-ligand binding affinities have long-established prominence, binding rate constants and binding mechanisms have gained increasing attention in recent years. Both new computational methods and new experimental techniques have been developed to characterize the latter properties. It is now realized that binding mechanisms, like binding rate constants, can and should be quantitatively determined. In this review, we summarize studies and synthesize ideas on several topics in the hope of providing a coherent picture of and physical insight into binding kinetics. The topics include microscopic formulation of the kinetic problem and its reduction to simple rate equations; computation of binding rate constants; quantitative determination of binding mechanisms; and elucidation of physical factors that control binding rate constants and mechanisms.

Protein Binding Pocket Dynamics

The dynamics of protein binding pockets are crucial for their interaction specificity. Structural flexibility allows proteins to adapt to their individual molecular binding partners and facilitates the binding process. This implies the necessity to consider protein internal motion in determining and predicting binding properties and in designing new binders. Although accounting for protein dynamics presents a challenge for computational approaches, it expands the structural and physicochemical space for compound design and thus offers the prospect of improved binding specificity and selectivity. A cavity on the surface or in the interior of a protein that possesses suitable properties for binding a ligand is usually referred to as a binding pocket. The set of amino acid residues around a binding pocket determines its physicochemical characteristics and, together with its shape and location in a protein, defines its functionality. Residues outside the binding site can also have a long-range effect on the properties of the binding pocket. Cavities with similar functionalities are often conserved across protein families. For example, enzyme active sites are usually concave surfaces that present amino acid residues in a suitable configuration for binding low molecular weight compounds. Macromolecular binding pockets, on the other hand, are located on the protein surface and are often shallower. The mobility of proteins allows the opening, closing, and adaptation of binding pockets to regulate binding processes and specific protein functionalities. For example, channels and tunnels can exist permanently or transiently to transport compounds to and from a binding site. The influence of protein flexibility on binding pockets can vary from small changes to an already existent pocket to the formation of a completely new pocket. Here, we review recent developments in computational methods to detect and define binding pockets and to study pocket dynamics. We introduce five different classes of protein pocket dynamics: (1) appearance/disappearance of a subpocket in an existing pocket; (2) appearance/disappearance of an adjacent pocket on the protein surface in the direct vicinity of an already existing pocket; (3) pocket breathing, which may be caused by side-chain fluctuations or backbone or interdomain vibrational motion; (4) opening/closing of a channel or tunnel, connecting a pocket inside the protein with solvent, including lid motion; and (5) the appearance/disappearance of an allosteric pocket at a site on a protein distinct from an already existing pocket with binding of a ligand to the allosteric binding site affecting the original pocket. We suggest that the class of pocket dynamics, as well as the type and extent of protein motion affecting the binding pocket, should be factors considered in choosing the most appropriate computational approach to study a given binding pocket. Furthermore, we examine the relationship between pocket dynamics classes and induced fit, conformational selection, and gating models of ligand binding on binding kinetics and thermodynamics. We discuss the implications of protein binding pocket dynamics for drug design and conclude with potential future directions for computational analysis of protein binding pocket dynamics.