The ligand binding mechanism to purine nucleoside phosphorylase elucidated via molecular dynamics and machine learning

Abstract

The study of biomolecular interactions between a drug and its biological target is of paramount importance for the design of novel bioactive compounds. In this paper, we report on the use of molecular dynamics simulations and machine learning to study the binding mechanism of a transition state analogue (DADMe-Immucillin-H) to the purine nucleoside phosphorylase enzyme. Microsecond-long molecular dynamics simulations allow us to observe several binding events, following different dynamical routes and reaching diverse binding configurations. These simulations are used to estimate kinetic and thermodynamic quantities, such as kon and binding free energy, obtaining a good agreement with available experimental data. In addition, we advance a hypothesis for the slow-onset inhibition mechanism of DADMe-Immucillin-H against purine nucleoside phosphorylase. Combining extensive molecular dynamics simulations with machine learning algorithms could therefore be a fruitful approach to capturing key aspects of drug-target recognition and binding.

Development of a multisite model for Ni(II) ion in solution from thermodynamic and kinetic data

“…WTmetaD was performed with a bias factor of 15 at a temperature of 300 K. Gaussians with initial height of 0.05 kcal mol −1 were added every 1 ps, and the adaptive Gaussian width scheme was used . In this way, the Gaussian width is adapted on the fly to the local free energy surface to increase the sampling efficiency.…”

Computational chemical biology and drug design: Facilitating protein structure, function, and modulation studies

“…More recently, increased computational power combined with enhanced sampling methods and machine learning (ML) algorithms have allowed researchers to observe the ligand-binding mechanism more comprehensively. Decherchi et al 66 have devised an automated pipeline combining ML and MD simulations to study the binding mechanism of a transition state analogue (DADMe-immucillin-H) to the purine nucleoside phosphorylase enzyme. They have generated microsecond-long MD simulations and used the k-medoids algorithm to cluster all the trajectories and capture key binding events.…”

Understanding ligand-receptor non-covalent binding kinetics using molecular modeling

“…This results in the shortcomings of the conventional MD method, which limits the timescale that it can practically achieve. Without accelerating techniques, the timescale that conventional MD can reach is limited to microseconds with contemporary hardware (22, 24–26) leaving a large gap to the timescale of protein motion of interest. However despite the limitations of timescale of conventional molecular dynamics, there have been important discoveries and progress over the past years.…”

“…This enabled study of the complete drug binding process of a transition state analogue to purine nucleoside phosphorylase (PNP) by microsecond level MD simulations with the aid of machine learning (22) on GPUs (76). Peptide recognition by protein domain was studied by microsecond MD simulation with parallel computing (77).…”

“…High energy barriers to binding and rough free energy surfaces naturally contribute to slow association/dissociation rates, which sometimes may be on the scale of seconds or even days. Both of these assert challenges to contemporary computational power in sampling the dynamics of drug-protein binding, which can only simulate conformational changes in the microsecond range (22, 24–26). Furthermore, even if the conformational changes are sampled, the extraction of kinetics, or probability of state transitions, remains as the second obstacle before rationalized design of drugs.…”