Efficient Sampling of High-Dimensional
Free Energy Landscapes: A Review
of Parallel Bias Metadynamics
Sarah Alamdari, Janani Sampath, Arushi Prakash, Luke D. Gibson,
and Jim Pfaendtner
1 Introduction
Molecular dynamics (MD) and Monte Carlo simulations are powerful methods used
to elucidate the conformational and structural properties of molecules by the dynamic
or stochastic evolution of a molecular configuration. Simulations find value across
a number of fields and range from understanding the fundamental driving forces in
a simple Lennard Jones system for gaining thermodynamic and kinetic insight of
complex biological and chemical systems. MD simulations, while a powerful tool
for sampling the dynamic evolution of a system, usually fail to sufficiently sample all
the relevant configurations of these complex systems. Further, when the interesting
states of the system are separated by free energy barriers larger than k B T, it becomes
impossible to comprehensively sample a 3 N-dimensional phase space. In order to
improve sampling in such simulations, a number of enhanced sampling methods
have been proposed. These methods have been successfully applied to study rare
events in biological processes (e.g., protein unfolding [1–3], peptide adsorption [4–
6], drug and ligand docking [7–9], phase transitions [10, 11], and nucleation [12,
13]) and chemical reactions [14–19] in combustion, pyrolysis, and enzymology to
name a few. This review is not meant to be comprehensive, but an overview of the
development of metadynamics and a few of its variants. Notable alternative enhanced
sampling methods including, but not limited to, transition path sampling [20], log
mean force dynamics [21], multi-scale enhanced sampling [22], average force [23],
nudged elastic band [24], Onsager-Machlup multi-scale enhanced sampling (MSES)
[25], boxed molecular dynamics (BXD) [26], blue moon sampling [27], and string
method [28] are outside of the scope of this chapter. These methods and more can
be found in excellent detail in various reviews [29–32].
S. Alamdari · J. Sampath · A. Prakash · L. D. Gibson · J. Pfaendtner (B)
Department of Chemical Engineering, University of Washington, Seattle, WA 98195, USA
e-mail: jpfaendt@uw.edu
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021
E. J. Maginn and J. Errington (eds.), Foundations of Molecular Modeling
and Simulation, Molecular Modeling and Simulation,
https://doi.org/10.1007/978-981-33-6639-8_6
123
Free Energy Landscapes: A Review
of Parallel Bias Metadynamics
Sarah Alamdari, Janani Sampath, Arushi Prakash, Luke D. Gibson,
and Jim Pfaendtner
1 Introduction
Molecular dynamics (MD) and Monte Carlo simulations are powerful methods used
to elucidate the conformational and structural properties of molecules by the dynamic
or stochastic evolution of a molecular configuration. Simulations find value across
a number of fields and range from understanding the fundamental driving forces in
a simple Lennard Jones system for gaining thermodynamic and kinetic insight of
complex biological and chemical systems. MD simulations, while a powerful tool
for sampling the dynamic evolution of a system, usually fail to sufficiently sample all
the relevant configurations of these complex systems. Further, when the interesting
states of the system are separated by free energy barriers larger than k B T, it becomes
impossible to comprehensively sample a 3 N-dimensional phase space. In order to
improve sampling in such simulations, a number of enhanced sampling methods
have been proposed. These methods have been successfully applied to study rare
events in biological processes (e.g., protein unfolding [1–3], peptide adsorption [4–
6], drug and ligand docking [7–9], phase transitions [10, 11], and nucleation [12,
13]) and chemical reactions [14–19] in combustion, pyrolysis, and enzymology to
name a few. This review is not meant to be comprehensive, but an overview of the
development of metadynamics and a few of its variants. Notable alternative enhanced
sampling methods including, but not limited to, transition path sampling [20], log
mean force dynamics [21], multi-scale enhanced sampling [22], average force [23],
nudged elastic band [24], Onsager-Machlup multi-scale enhanced sampling (MSES)
[25], boxed molecular dynamics (BXD) [26], blue moon sampling [27], and string
method [28] are outside of the scope of this chapter. These methods and more can
be found in excellent detail in various reviews [29–32].
S. Alamdari · J. Sampath · A. Prakash · L. D. Gibson · J. Pfaendtner (B)
Department of Chemical Engineering, University of Washington, Seattle, WA 98195, USA
e-mail: jpfaendt@uw.edu
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021
E. J. Maginn and J. Errington (eds.), Foundations of Molecular Modeling
and Simulation, Molecular Modeling and Simulation,
https://doi.org/10.1007/978-981-33-6639-8_6
123