Predicting Protein–Ligand Binding and Unbinding Kinetics with Biased MD Simulations and Coarse-Graining of Dynamics: Current State and Challenges

Abstract

The prediction of drug−target binding and unbinding kinetics that occur on time scales between milliseconds and several hours is a prime challenge for biased molecular dynamics simulation approaches. This Perspective gives a concise summary of the theory and the current state-of-the-art of such predictions via biased simulations, of insights into the molecular mechanisms defining binding and unbinding kinetics as well as of the extraordinary challenges predictions of ligand kinetics pose in comparison to binding free energy predictions.

Investigation of Rare Protein Conformational Transitions via Dissipation-Corrected Targeted Molecular Dynamics

“…Indeed, Figure S5c,d indicates a reasonably converged free energy estimate from pulling simulations. We note that correctly reproducing the T4L free energy profile poses a challenge for computational predictions of free energies, as the typical mean error of free energy calculation methods , is on the order of ∼3 k B T . Furthermore, the 1000 pulling simulations collected for T4L correspond to an accumulated ∼60 μs of simulated time, which is on the same order as our reference unbiased simulation but can be calculated in parallel and represent 2 orders of magnitude more transition events.…”

“…This previous study utilized the distance between the center of mass of the Phe4 side chain and the combined center of mass of residues Lys60 and Glu64 (Figure b) as the pulling coordinate s . Note that pulling the Phe4 side chain out of its hydrophobically buried position into the solvent is the only practical option to enforce the open–close conformational transition of T4L: the reverse reaction, i.e., pulling the side chain into the protein, requires establishing the correct packing near order, which is beyond the time scales desired in our biased simulations and is well understood in the case of protein–ligand complexes . We use this conformational transition as a prime example of the construction of a Langevin model via dcTMD.…”

“…To resolve this issue, enhanced sampling techniques may be employed to accelerate these rare transitions and thus estimate G and Γ . Popular examples that allow for the estimation of process rates are random acceleration MD, , weighted ensemble and milestoning approaches, infrequent metadynamics, , or Gaussian-accelerated MD. , In this work, we focus on dissipation-corrected targeted MD (dcTMD), which utilizes the idea of TMD by constraining the system along a suitable reaction coordinate s with constant velocity v , i.e. s ( t ) = s 0 + v t in order to enforce the transition between two states of interest. By calculating the work W from the applied constraint force f W ( s ) = ∫ 0 s normald s ′ f ( s ′ ) the free energy can be estimated from Jarzynski’s equation .25ex2ex normalΔ G ( s ) = prefix− β − 1 ln nobreak0em0.25em⁡ ⟨ e − β W false| s ⟩ false( 4 normala false) ≈ ⟨ W | s ⟩ − β 2 ⟨ δ W 2 | s false⟩ false( 4 normalb false) …”

Kinetics from Metadynamics: Principles, Applications, and Outlook

“…In the field of biology, most applications of enhanced sampling approaches for kinetics calculation have focused on the binding/unbinding kinetics of protein–ligand complexes. − This can be attributed to the increased realization that ligand residence time is a key predictor of drug efficacy. − Therefore, efforts have been dedicated to the development and application of computational methods to predict the drug–receptor unbinding rate constant ( k off ). In Table , we summarize the protocol and results of such studies involving metadynamics-based approaches.…”

Manifold learning in atomistic simulations: a conceptual review

“…As a result, the system remains kinetically trapped in a metastable state as its dynamics is restricted to sampling fast equilibration. This so-called sampling problem can be observed in many physical processes, for example, catalysis [66], ligand interactions with proteins [67][68][69][70][71] and DNA [72], glass transitions in amorphous materials [73], crystallization [46], and graphite etching [74]. (zl) .…”