pubs.acs.

org/JCTC Article

py-MCMD: Python Software for Performing Hybrid Monte Carlo/

Molecular Dynamics Simulations with GOMC and NAMD
Mohammad Soroush Barhaghi, Brad Crawford, Gregory Schwing, David J. Hardy, John E. Stone,
Loren Schwiebert, Jeffrey Potoff,* and Emad Tajkhorshid*
Cite This: J. Chem. Theory Comput. 2022, 18, 4983−4994 Read Online

ACCESS Metrics & More Article Recommendations *

sı Supporting Information
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

ABSTRACT: py-MCMD, an open-source Python software, provides a

robust workflow layer that manages communication of relevant system
Downloaded via UNIV OF DELAWARE on December 4, 2024 at 17:51:19 (UTC).

information between the simulation engines NAMD and GOMC and

generates coherent thermodynamic properties and trajectories for analysis.
To validate the workflow and highlight its capabilities, hybrid Monte Carlo/
molecular dynamics (MC/MD) simulations are performed for SPC/E water
in the isobaric−isothermal (NPT) and grand canonical (GC) ensembles as
well as with Gibbs ensemble Monte Carlo (GEMC). The hybrid MC/MD
approach shows close agreement with reference MC simulations and has a
computational efficiency that is 2 to 136 times greater than traditional Monte
Carlo simulations. MC/MD simulations performed for water in a graphene
slit pore illustrate significant gains in sampling efficiency when the coupled−decoupled configurational-bias MC (CD−CBMC)
algorithm is used compared with simulations using a single unbiased random trial position. Simulations using CD−CBMC reach
equilibrium with 25 times fewer cycles than simulations using a single unbiased random trial position, with a small increase in
computational cost. In a more challenging application, hybrid grand canonical Monte Carlo/molecular dynamics (GCMC/MD)
simulations are used to hydrate a buried binding pocket in bovine pancreatic trypsin inhibitor. Water occupancies produced by
GCMC/MD simulations are in close agreement with crystallographically identified positions, and GCMC/MD simulations have a
computational efficiency that is 5 times better than MD simulations. py-MCMD is available on GitHub at https://github.com/
GOMC-WSU/py-MCMD.

■ INTRODUCTION
Molecular dynamics (MD) simulations are an essential tool for
These systems contain high free energy barriers, which may
prevent reliable sampling of phase space during conventional
understanding biological structure and function. Simulations of MD simulation time scales, or require the use of an open
systems containing 100 000 atoms are routine, and simulations system, i.e., allowing for changes in the number of molecules.
of up to 2 billion atoms have been made possible through An alternative strategy to running longer simulations is to
advances in hardware and MD codes.1 For moderate-sized incorporate Monte Carlo (MC) moves into MD simula-
systems, microsecond and even millisecond time scales are tions.17,19−21,23−26 In this hybrid approach, trial MC moves are
attainable nowadays.2,3 MD simulations are used routinely for proposed, which are accepted or rejected on the basis of
computational drug discovery,4−6 determination of lipid statistical mechanical probabilities. These MC moves may use
bilayer mechanical and transport properties,2,7−10 and alternative, or “unphysical”, pathways that allow the system to
elucidation of protein function.11−13 Since all atomic positions traverse high free energy barriers. The inclusion of MC moves
are known at any moment in time, MD simulations provide also provides the opportunity to simulate systems in the grand
spatial and temporal resolutions that are not currently canonical (GC) ensemble, where the number of constituent
achievable with experiments, and therefore, the method can molecules can change during the simulation.
The use of grand canonical Monte Carlo (GCMC) sampling
be viewed as a “computational microscope”14,15 in the quest to
in MD simulations has been shown to produce significant
understand biological machinery.
While the MD methodology is broadly applicable to the
study of biological molecules, there are certain problems that Received: September 10, 2021
may be better suited to alternative sampling approaches. Published: May 27, 2022
Notable examples include diffusion through pores and
membranes,16 the hydration state of buried pockets or
channels in proteins,17−21 the formation of nanodomains,3
and phase separation in multicomponent lipid bilayers.22,23

© 2022 American Chemical Society https://doi.org/10.1021/acs.jctc.1c00911

4983 J. Chem. Theory Comput. 2022, 18, 4983−4994
Journal of Chemical Theory and Computation pubs.acs.org/JCTC Article

improvements in the accuracy and precision of ligand−protein

binding free energies.20,21,24,27,28 GCMC sampling for water in
the binding pocket leads to rapid equilibration, compared with
the slower diffusion observed in MD, particularly for buried
sites. GCMC, using dual control volumes,16 has been
combined with Brownian dynamics (GCMC−BD) to simulate
the movement of ions through transmembrane pores.29,30 The
inclusion of an MC identity-exchange move, using either
configurational-bias sampling22 or a combination of alchemical
transformation and nonequilibrium MD,23 to sample phase
space has been used to greatly accelerate the equilibration of
multicomponent lipid bilayers.
Basic GC functionality is present in a number of MD
codes17,19,25,26,31 in a variety of forms: either integrated into
the code, as has been done in CHARMM,17,31 AMBER,19 and
LAMMPS,32 or interfaced through external Python codes for
GROMACS25 and OPENMM.26 Additional MC moves are
available in OpenMM33 through the OpenMMTools pack-
age.34 Some implementations use the cavity-bias method to
improve acceptance rates for the transfer of water molecules,35
while in others only random insertions are possible.
This work describes hybrid MC/MD simulations that utilize
NAMD for MD and GOMC for MC. These codes are linked
with py-MCMD, an open-source Python program that
oversees information transfer and the execution of each
code.36 By linking of GOMC with NAMD, it is possible to
perform hybrid GCMC/MD simulations that utilize the full
suite of advanced configurational-bias sampling algorithms that
are available in GOMC.37−40 With py-MCMD, it is also
possible to integrate MD sampling of configurational space
into MC simulations, such as Gibbs ensemble (GE),41 leading
to enhanced sampling efficiency over standard Gibbs ensemble
Monte Carlo (GEMC) simulations.42
To validate the py-MCMD program and assess its
computational efficiency, a series of MC/MD simulations
were performed for water in the isothermal−isobaric (NPT) Figure 1. Schematic of the hybrid MC/MD workflow. In the initial
cycle, MD simulations with or without restraints are used to minimize
and GC ensembles as well as with the GE method. Additional the energy of the system and perform a short dynamics simulation to
hybrid MC/MD simulations were performed in the canonical eliminate unfavorable configurations. This is followed by alternating
(NVT) ensemble with an intrabox swap move for water in a cycles of MC and MD simulations. During the MC simulations, a
graphene slit pore and in the GC ensemble to determine the variety of MC moves are utilized in each ensemble to satisfy
hydration state of a buried binding site in bovine pancreatic equilibrium conditions, e.g., swap (transfer of a molecule between
trypsin inhibitor (BPTI). Notably, NVT MC/MD simulations phases), intraswap (deletion and reinsertion of a molecule in a
of water in a graphene slit pore, performed with coupled− random new location in the same phase), or volume change
decoupled configurational-bias MC (CD−CBMC), equili- (perturbing the volume of the system). MD simulations in the
brated in 25 times fewer cycles than simulations without NVT ensemble are used for efficient sampling of configurations and
CD−CBMC and 66 times fewer cycles than hybrid MC/MD conformations. Control of the simulation parameters, ensemble, and
MC moves are possible through NAMD and GOMC control files
simulations using the cavity-bias method.19 GCMC/MD (hexagons). Transfer of information between the MC and MD
simulations of the BPTI system showed rapid hydration of engines is overseen by the py-MCMD Python software.
the buried binding pocket with a 5-fold improvement in
computational efficiency compared with standard MD
simulations. disk space requirements. These modifications are available in

■ METHODS
Workflow. py-MCMD defines a hybrid MC/MD workflow
the recently released GOMC version 2.75. py-MCMD was
designed to work with NAMD version 2.14. The py-MCMD
software and its documentation are available on GitHub at
that facilitates information transfer between GOMC and https://github.com/GOMC-WSU/py-MCMD.
NAMD (Figure 1).36 An additional Python analysis script The hybrid MC/MD workflow starts with a short conjugate-
was also created that combines the simulation trajectories and gradient energy minimization and NVT MD simulation to
thermodynamic properties reported by NAMD and GOMC stabilize the system, which is followed by cycles of alternating
into several compact files for postsimulation analysis and MC and MD steps. Following the strategy of Gartner et al., the
visualization. GOMC version 2.7040 was modified to read and results of the MD simulation are always accepted,42 while all
write binary coordinate, velocity, and extended system control MC moves follow strict detailed balance. The choice to use
files in NAMD native format, which improves the accuracy of nonmetropolized hybrid Monte Carlo was based on the
the hybrid simulations and reduces simulation startup time and following criteria: it produces correct results for reasonable
4984 https://doi.org/10.1021/acs.jctc.1c00911
J. Chem. Theory Comput. 2022, 18, 4983−4994
Journal of Chemical Theory and Computation pubs.acs.org/JCTC Article

choices of simulation parameters; it is computationally more

2000 MCS + 10 ps
20 MCS + 1.96 ps

200 MCS + 2.6 ps

efficient than metropolized hybrid Monte Carlo;43 and the

7000 MCS + 6 ps
1000 MCS + 2 ps
1000 MCS + 2 ps
500 MCS + 2 ps

500 MCS + 2 ps
output freq.
algorithm is straightforward to implement. It should be noted,

10000 MCS
1000 MCS

1500 MCS

1500 MCS

1500 MCS
however, that using the Brooks−Brünger−Karplus (BBK)
integrator44 in Langevin dynamics simulations with a large

10 ps
time step will introduce a time-step-dependent bias45 due to
finite integration time step error.43,46 Despite this bias, the
BBK integrator does produce a configuration-space average

no. of replicas
that is close to the true ensemble average. In order to more

simulation parameters
accurately sample the exact configuration space, the BBK

5
5
5
5
5
5
5
5
5
5
5
5
10
10
integrator should be used with an integration time step of ≤1
fs, or alternatively, the BAOAB integrator47 can be used for
better conservation of the configuration-space average with a 2

cycle (MCS: ps)

fs time step.45,46

(2000: 10.0)
(7000: 6.0)
(1000: 2.0)
(1000: 2.0)
(20: 1.96)

(500: 2.0)

(500: 2.0)

(200: 2.6)
py-MCMD generates control files for GOMC and NAMD,
sets run-time parameters, launches the appropriate simulation

N/A

N/A

N/A

N/A

N/A

N/A
engine at each point in the sequence, and checks the total
system energies for consistency when switching between
engines. Relative energy differences between NAMD and

1.5 × 108 MCS

GOMC greater than 1 × 10−3 produce a warning message.

3 × 107 MCS

5 × 107 MCS

5 × 107 MCS

5 × 107 MCS
a
run length

1.3 × 104 C

1.3 × 104 C

1.3 × 104 C
Typical relative energy differences observed between GOMC

3 × 104 C

1 × 105 C

2 × 104 C
2 × 104 C

1 × 103 C
and NAMD were on the order of 1 × 10−6 to 1 × 10−5. To

10 ns
maintain continuity of system dynamics, velocity information is
passed through GOMC, even though it is not required for MC
simulations. Any atoms that are moved or inserted during the

regrowth
Monte Carlo phase have their velocities initialized using a

0.25

0.20

0.20

0.20

0.20
Maxwell−Boltzmann distribution.48,49
All of the simulations were performed using a switched
potential for Lennard−Jones (LJ) interactions using a 12 Å
rotate
0.25

0.15

0.20

0.20

0.20
cutoff and a 10 Å switch distance. No long-range corrections
were used for LJ interactions. Electrostatic interactions were
distribution of MC moves

calculated using the Ewald summation50,51 in GOMC and

translate
0.25

0.15

0.19

0.19

0.19
particle-mesh Ewald52,53 in NAMD. A tolerance of 10−5 and a
real-space cutoff of 12 Å were used for electrostatic
interactions. Molecule-transfer and intrabox swap moves
intraswap

were performed with the CD−CBMC method37 using 16

0.24

0.20

0.20

0.20

1.00
1.00
trial positions for the first atom and eight trial positions for all
remaining atoms.54 A derivation of the CD−CBMC algorithm
for flexible and rigid-body molecules is provided in the

1.00b
swap

0.50
1.00
0.20
0.99
0.20
0.99
0.20
0.99

Supporting Information. A hard inner cutoff was used to reject

any MC moves that placed atom centers closer than 1 Å.55 In
both MC and MD simulations, water was treated as a rigid-
volume
0.01
1.00

0.01
0.01
0.01
0.01
0.01
0.01

body molecule, with covalent bonds to hydrogen atoms

constrained using the SETTLE algorithm56 in the MD
simulations. All of the simulations were performed using
ensemble [T (K)]

periodic boundary conditions with the minimum image

GCMC [510]
GCMC [510]

GCMC [298]
GEMC [500]
GEMC [500]
GEMC [400]
GEMC [400]
GEMC [300]
GEMC [300]
NVT [500]
NVT [500]
NVT [298]
NPT [298]
NPT [298]

convention. MD simulations were performed in the NVT

Table 1. Summary of Simulation Parameters

MCS, MC steps; C, cycles. bTargeted swap move.

ensemble using Langevin dynamics57 with the BBK integrator

scheme44 to simulate the heat bath. For the MD simulations,
the multiple-time-step r-RESPA algorithm58 was used with a 2
fs integration time step, where short- and long-range forces
were evaluated every 2 and 4 fs, respectively.
MC/MD

MC/MD

MC/MD

MC/MD

MC/MD
MC/MD
MC/MD

MC/MD
simulation type
type

Workflow Validation: SPC/E Water. To validate the

MD
MC

MC

MC

MC

MC

workflow, hybrid MC/MD simulations were performed for

SPC/E water59 in the NPT, GC, and Gibbs ensembles, and the
results were compared to reference MC simulations. A variety
slit pore (CD−CBMC)

of Monte Carlo moves were used, which include rigid-body

slit pore (unbiased)

translation (translate), rigid-body rotation (rotate), volume

system

exchange (volume), molecule transfer between phases (swap),

SPC/E water

and configurational-bias regrowth (regrowth). A summary of

the MC move fractions and lengths of the MC and MD
BPTI

components for each system is provided in Table 1. A

description of the MC moves used in this work, such as rigid-
a

4985 https://doi.org/10.1021/acs.jctc.1c00911
J. Chem. Theory Comput. 2022, 18, 4983−4994
Journal of Chemical Theory and Computation pubs.acs.org/JCTC Article

body displacement and rotation, configurational bias, and using standard MD simulations is grossly inefficient or in some
volume exchange are provided in the Supporting Information. cases impossible (e.g., graphene slit pore equilibration). The
Parameters for the SPC/E water model are listed in Table S1. NVT MC/MD simulations at 500 K were performed on SPC/
For all of the hybrid MC/MD simulations, the numbers of MC E water in a rectangular box with dimensions of 29.5 Å × 29.8
trials and MD time steps were set to ensure that each cycle Å × 85 Å that was partitioned into two regions by two
would result in at least one accepted MC move and one impermeable graphene sheets placed at 28 Å apart from each
uncorrelated sample during the MD simulation. To calculate other.19 Each graphene sheet contained 336 carbon atoms.
statistical uncertainties, each simulation was repeated five Force field parameters for carbon atoms are listed in Table
times, initiated with different random seeds and with unique S1.67 The outer region was filled initially with 1866 water
initial configurations (coordinates and velocities) generated molecules at a density of 1.197 g/cm3, while the inner region
using the Molecular Simulation Design Framework (MoS- was populated with a single water molecule. To compare the
DeF).60−63 efficiency of the CD−CBMC algorithm with simple MC
Simulations in the NPT ensemble were performed at 298 K sampling, additional simulations were performed using only
and 1.01325 bar for a system containing 1361 SPC/E water one unbiased trial position for the intrabox swap move. Five
molecules with an initial density of 0.9496 g/cm3. GC independent NVT MC/MD simulations were performed to
ensemble simulations were performed at 510 K and a chemical calculate statistical uncertainties, where each simulation was
potential of −4.9037 kcal/mol, with a box size of V̅ = 42 875 initiated with different random seeds, unique initial config-
Å3. GE simulations41 were performed at 300, 400, and 500 K to urations, and velocities.
calculate the saturated liquid and vapor coexistence densities of In the final example, GCMC/MD simulations were used to
SPC/E water. For the GE simulations at 500 K, the initial calculate the most probable water positions in the buried
configurations for the vapor and liquid phases each included binding pocket of BPTI. The initial configuration was
500 molecules at a density of 0.295 g/cm3. The initial density generated from the starting structure (PDB ID 5PTI),68−71
of each phase was intentionally set far from equilibrium to without any hydrogen/deuterium atoms. The system was
verify the convergence of the MC/MD simulation. Initial protonated and solvated in water with 15 Å of padding around
densities for simulations at 300 and 400 K were set to the the protein in each dimension, and neutralizing ions (0.15
expected saturated liquid and vapor densities, generated from mol/L NaCl) were added to the system using the QwikMD72
prior simulations of SPC/E water.64 plugin in VMD.73 The mTIP3P74 and CHARMM3675−83 force
Computational Efficiency. To calculate the computa- fields were used to represent water and protein interactions,
tional efficiency of the hybrid MC/MD with respect to respectively. The initial configuration was minimized for 2 ps
standard MC simulations, the integrated autocorrelation time and then annealed with NPT MD simulations from 60 to 298
(τ), statistical inefficiency (g = 1 + 2τ), and normalized K for 28.68 ps, followed by a 5 ns NPT MD equilibration at
fluctuation autocorrelation function (C(t)) observed during 298 K and 1.01325 bar. The Langevin piston method84 was
each simulation were calculated using the pymbar autocorre- used for the barostat, which combines the Hoover constant-
lation time analysis tool.65,66 To accurately compare the pressure equations of motion85,86 with piston fluctuations
computational efficiencies of the hybrid MC/MD and MC controlled by Langevin dynamics.84 During the equilibration,
simulations, τ, g, and C(t) are reported in terms of the protein’s backbone atoms were restrained using a harmonic
computational cost (CPU hours) instead of number of potential with a 2 kcal mol−1 Å−2 force constant.
simulation steps. g = 1 indicates that 1 CPU hour was GCMC/MD simulations were performed at 298 K and a
required to generate an uncorrelated data point. All of the bulk chemical potential of −6.3349 kcal/mol. This corresponds to
water simulations were performed on four cores of an Intel the average equilibrium density of 1.0113 ± 0.0004 g/cm3 for
Gold 5118 2.3 GHz CPU. Data points were written to disk bulk mTIP3P water, which is consistent with the density
with the frequencies listed in Table 1. produced by NPT MD simulations at 298 K and 1.01325 bar.87
Data series were generated for autocorrelation analysis after Water molecules were inserted and deleted from a cube with a
equilibration was reached according to the reduced potential side length of 15 Å centered at the geometric center of the Cα
function f(R) appropriate to each ensemble.65 The f(R) atoms of the Tyr10 and Asn43 residues.26 For these
function for a particular microstate is given by calculations, a hard inner cutoff for MC insertion/deletion
f (R) = [U (R) + PV (R) N (R)] moves was not used since all of the atoms in the mTIP3P water
(1)
force field have Lennard−Jones parameters, minimizing the
where R, U(R), V(R), and N(R) are the configuration, total possibility of naked charges being placed in close proximity
potential energy, volume, and number of molecules in the during an insertion attempt. Ten independent GCMC/MD
system for a specific microstate, respectively, β = 1/kBT, P is simulations were performed for 1000 cycles, where each cycle
the imposed pressure, and μ is the imposed chemical potential. consisted of 2000 MC steps and a 10 ps MD simulation. Each
Depending on the ensemble, eq 1 combines temperature, simulation was started with an empty binding pocket and a
potential energy, pressure, volume, number of molecules, and/ unique initial configuration, distribution of velocities, and
or chemical potential to produce a reduced-value data series. random number seed. To prevent collapse of the binding
The reduced potential functions used in this work for the NPT pocket due to the absence of water, additional harmonic
and μVT ensembles were −β[U(R) + PV(R)] and −β[U(R) − restraints with a 2 kcal mol−1 Å−2 force constant were applied
μN(R)], respectively. For GEMC simulations, the reduced to heavy atoms of the protein during the first cycle of the
potential used was f(R) = −β[Uliq(R) + Ugas(R)]. GCMC/MD simulations. The rest of the cycles were run
Illustrative Applications: Graphene Slit Pore and without restraints.
BPTI. To illustrate the application of the py-MCMD software, To determine the most probable positions for each of the
two additional hybrid MC/MD simulations were performed in three water molecules within the binding pocket, the
the NVT and GC ensembles, where achieving equilibrium simulation trajectories were aligned, and water molecule
4986 https://doi.org/10.1021/acs.jctc.1c00911
J. Chem. Theory Comput. 2022, 18, 4983−4994
Journal of Chemical Theory and Computation pubs.acs.org/JCTC Article

positions observed during the simulation were clustered across are shown in Figure S1. Analysis of the correlation time τ for
trajectory frames using the average-linkage hierarchical −β(U + PV) for MC/MD and MC simulations (Table 2)
clustering algorithm, as implemented by Samways et al.26 A revealed that the correlation time was 136 ± 58 times lower for
water molecule was considered to be clustered if its oxygen MC/MD simulations than for standard MC.
atom was within 2.4 Å of another water oxygen atom. To avoid GCMC/MD simulations were performed for water at 510 K
clustering of waters within the same frame, the oxygen−oxygen and a chemical potential of −4.9037 kcal/mol, which produced
distance of waters that belonged to the same frame in a an average density of 0.7839 ± 0.0006 g/cm3, compared to an
trajectory was set to a large number (∼108 Å). Each cluster average density of 0.786 ± 0.001 g/cm 3 for GCMC
consisted of an array of oxygen positions, and the cluster simulations. This state corresponds to a density slightly higher
centroid was calculated as the average of these oxygen than the saturated liquid density for SPC/E water of 0.771 g/
positions. The most probable water positions were calculated cm3 at 510 K.64,88,89 As shown in Figure 2, the GCMC/MD
on the basis of the oxygen position of the water molecule that simulation trajectories closely follow those of GCMC.
is observed closest to the cluster centroid. The occupancy of Autocorrelation functions for −β(U − μN) are shown in
clustered water was calculated as the ratio of total observed Figure S2, and results for the correlation time analysis are
waters in the cluster to the number of frames in the trajectory. presented in Table 2. The GCMC/MD approach decreased
Clustered water positions were assigned to a specific the correlation time by a factor of 2.1 ± 0.4 compared with the
crystallographic water site (e.g., site 1, 2, or 3) if they were MC simulations. These results are not as dramatic as those
within 1.5 Å of one crystallographic water site and not closer to from the NPT simulations but are expected since sampling of
any others. −β(U − μN) is limited primarily by the molecule-transfer

■ RESULTS AND DISCUSSION

Workflow Validation: SPC/E Water. To validate the
move, and MD sampling does not seem to have a significant
impact on the percentage of accepted molecule-transfer
attempts.
hybrid MC/MD workflow, SPC/E water simulations were The corresponding probability distribution for the number
performed in the isobaric−isothermal (NPT), and grand of water molecules in the system is presented in Figure 3,
canonical (GC) ensembles. Calculations were also performed where the GCMC/MD data show a subtle shift to lower
with the Gibbs ensemble (GE) Monte Carlo method. As density (∼0.2%) compared with the GCMC simulations.
shown in Figure 2, close agreement was achieved between the Using the BBK integrator44 in Langevin dynamics simulations
trajectories generated with MC/MD and standard MC with a 2 fs time step results in overheating of the system by
simulations. approximately 1 K and a corresponding shift in the potential
energy distribution.45 This leads to sampling of configuration
space at a slightly higher configurational temperature and an
underprediction of the density in GCMC/MD simulations. If
more accurate sampling of configuration space is required, a
time step of ≤1 fs should be used with the BBK integrator.
Alternatively, a time step of up to 2 fs may be used with the
BAOAB integrator,47 which has been shown to better conserve
sampling of configuration space with larger time steps.43,45
Gibbs ensemble MC provides a direct means for the
calculation of vapor−liquid equilibria.41 In this method, two
phases (liquid and vapor in this case) are simulated
simultaneously. Equilibrium between the phases is achieved
through MC moves that exchange volume and molecules
between the phases, ensuring mechanical and chemical
equilibrium. It has been suggested that the efficiency of the
GEMC method could be improved by using an MC/MD
approach (GEMC/MD), where MC moves are used for only
the exchange of volume and molecules between the two phases
Figure 2. Predictions of MC/MD (black) and MC (red) simulations while MD steps are used to sample conformational and
as functions of computational cost (CPU-hr) in various ensembles for configurational degrees of freedom.42,90
SPC/E water: NPT at 298 K and 1.01325 bar; GCMC at 510 K and a GEMC/MD simulations were used to determine the vapor−
chemical potential of −4.9037 kcal/mol; GEMC at 500 K liquid and liquid equilibria of SPC/E water at 300 K, 400 K, and 500 K.
vapor densities. Solid lines correspond to averages over five replicas, Simulations at 500 K were purposefully started far from
while the shaded areas represent the standard deviations calculated equilibrium to observe the convergence behavior of the
from the five independent replicas.
algorithm, which is shown in Figure 2. Similar data for the
400 and 300 K simulations are presented in Figure S3.
NPT simulations at 298 K and 1.01325 bar were purpose- Tabulated data for saturated liquid and vapor densities and
fully started from an artificially low density of 0.9496 g/cm3 to vapor pressures are listed in Table 2. GEMC/MD results are in
test the convergence behavior of the MC/MD simulations. close agreement with data generated from standard GEMC
Figure 2 shows that the MC/MD simulations converged to the simulations. Both the GEMC/MD and GEMC simulations at
correct density (0.9939 ± 0.0001 g/cm3 for MC/MD 500 K converge to the correct equilibrium densities, with the
simulations and 0.995 ± 0.003 g/cm3 for MC simulations) GEMC/MD simulations reaching equilibrium faster than the
approximately 2 orders of magnitude faster than MC GEMC ones. Autocorrelation functions for −β(Uliq + Uvap) are
simulations. Autocorrelation functions C(t) for −β(U + PV) presented in Figures S4−S6 for simulations at 300, 400, and
4987 https://doi.org/10.1021/acs.jctc.1c00911
J. Chem. Theory Comput. 2022, 18, 4983−4994
Journal of Chemical Theory and Computation pubs.acs.org/JCTC Article

Table 2. Summary of Thermodynamic Properties and Correlation Analysis for MC/MD and Standard MC Simulations
Performed on SPC/E Watera
density (g/cm3) vapor pressure (bar) τ (CPU h) g (CPU h)
ensemble T (K) MC/MD MC MC/MD MC MC/MD MC MC/MD MC
NPT 298 0.9939(1) 0.995(3) − − 0.023(3) 3(1) 0.065(7) 6(3)
GCMC 510 0.7839(6) 0.786(1) − − 0.12(1) 0.26(5) 0.26(3) 0.52(9)
GEMC 500 0.7834(6) 0.787(4) − − 0.10(2) 0.4(1) 0.23(4) 0.8(2)
9.29(9) × 10−3 9.4(2) × 10−3 16.9(1) 16.9(2) − − − −
400 0.9135(4) 0.914(2) − − 0.6(3) 2(2) 1.3(5) 3(3)
6.8(1) × 10−4 6.7(1) × 10−4 1.16(1) 1.18(2) − − − −
300 0.993(1) 0.995(3) − − 1.8(5) 6(3) 4(1) 12(7)
7.6(2) × 10−6 7.8(5) × 10−6 0.0105(4) 0.0108(7) − − − −
a
Data correspond to averages taken over five replicas after the simulations reached equilibrium. Numbers in parentheses correspond to the
uncertainty in the last digit.

the correlation time by a factor of 3 to 4 compared with

standard GEMC simulations, with the greatest improvement
observed at 500 K. These results are consistent with the
GCMC/MD and GCMC simulations, as shown in Table 2,
where the molecule-transfer move is the primary limitation in
sampling efficiency. The smaller improvement of GEMC/MD
over GEMC at lower temperatures was expected, as the
percentage of accepted molecule-transfer moves also decreases
significantly under these conditions.
In summary, the MC/MD method produced uncorrelated
data significantly faster than the standard MC method, as
shown in Figures S1, S2, S4, S5, and S6. The normalized
fluctuation autocorrelation functions produced from MC/MD
simulations have significantly lower variance than those
generated from MC simulations. These results show that for
a given computational effort, more uncorrelated data were
obtained with MC/MD simulations than with MC simulations.
Figure 3. Probability distributions for the number of SPC/E water Illustrative Applications: Graphene Slit Pore and
molecules calculated with GCMC/MD using Langevin dynamics with BPTI. Now that the ability of the hybrid workflow to produce
a 2 fs time step (green) or a 1 fs time step (red) or standard GCMC
correct results for simulations of bulk water has been
simulations at 510 K and chemical potential of −4.9037 kcal/mol
(black). Solid lines correspond to averages over five independent demonstrated, two additional examples are presented to
replicas, while the shaded areas represent the standard deviations. highlight the utility of the py-MCMD program.
In Figure 4, data are presented for simulations of water in a
periodic rectangular box, partitioned into two separate
500 K, respectively. A correlation time analysis performed on compartments by two impermeable graphene sheets.19 In
−β(Uliq + Uvap) shows that the GEMC/MD approach reduced MD simulations, the inner region is not accessible because

Figure 4. NVT simulations of water in a graphene slit pore. (A) Initial configuration. (B) Final configuration after the MC/MD simulation was
performed. (C) Bulk water densities inside and outside the pore using CD−CBMC (black and blue) and simple MC (red and orange). Solid lines
correspond to averages over five replicas, and the shaded areas represent standard deviations calculated from the five independent replicas.

4988 https://doi.org/10.1021/acs.jctc.1c00911
J. Chem. Theory Comput. 2022, 18, 4983−4994
Journal of Chemical Theory and Computation pubs.acs.org/JCTC Article

Figure 5. Hydration of a buried binding pocket in BPTI. (A) Comparison of crystallographic water locations68 (cyan) with average water locations
predicted from GCMC/MD simulations (red). Any water positions that were observed less than 40% of the time have been omitted for clarity. (B)
Times required for GCMC/MD simulations (black) and MD simulations (red) to fill water sites in the binding pocket. The average occupancy for
each water site is shown as an average over water occupancies during 0.1 ns of simulation. The total number of water molecules in the pocket was
calculated by counting oxygen atoms within 4.2 Å of the geometric center of the Cα atoms of Tyr10 and Asn43. Solid lines correspond to averages
over 10 independent simulations, while the shaded areas represent the standard deviations. Data for each replica are presented in Figures S7−S10.

Figure 6. Effect of hydration on binding pocket stability in BPTI given by representative single MC/MD and MD simulations. (A) Comparison of
the crystallographic structure68 (purple) with the MD simulation (green). The green, yellow, and orange lines represent the top three random coil
(residues 36−47) clusters within the first 2 ns of MD simulation, which block the crystallographic water sites. (B) The 50 ps running averages of
RMSD calculations of the protein’s backbone for residues 36−47 with respect to the reference crystallographic structure68 for the MC/MD (black)
and MD (red) simulations. (C) The 50 ps running averages of the total number of observed waters within the binding pocket for the MC/MD
(black) and MD (red) simulations. Individual replicas are presented in Figure S11. The binding pocket is defined as a sphere of radius 4.2 Å
centered at the geometric center of the Cα atoms of Tyr10 and Asn43.

water cannot diffuse through the graphene layers; however, random trial position required approximately 18 750 cycles to
with the MC/MD approach and the intrabox swap moves, the reach equilibrium. In this application, the computational cost
entire simulation volume is accessible. Convergence of the of the CD−CBMC algorithm is approximately 25% larger than
density in each region was achieved in approximately 750 that of the simple MC algorithm.
cycles, where each cycle comprised 1000 CD−CBMC intrabox Both approaches used here compare favorably with prior
swap moves and a 2 ps MD simulation. Additional simulations simulations of this system using the cavity-bias method, which
were performed using an unbiased single random trial position required 50 000 cycles to reach equilibrium, where each cycle
(simple MC) for the intrabox swap move to illustrate the consisted of 25 000 MC trials and a 50 ps MD simulation.19
impact of the CD−CMBC method on the efficiency of the MC/MD simulations with CD−CBMC and simple MC
MC/MD simulation. Simulations using an unbiased single intrabox swap moves required 66 and 2.6 times fewer cycles
4989 https://doi.org/10.1021/acs.jctc.1c00911
J. Chem. Theory Comput. 2022, 18, 4983−4994
Journal of Chemical Theory and Computation pubs.acs.org/JCTC Article

to reach equilibrium, respectively. Additionally the cycles used binding pocket was dehydrated, and large fluctuations in the
in this work were 25 times shorter than those used by Ben- random coil were observed. After the binding pocket was fully
Shalom et al.,19 further illustrating the significant improvement hydrated, fluctuations in the RMSD decreased from 1.6 Å to an
in sampling achieved with the CD−CBMC algorithm. average of 0.7 Å. On the other hand, GCMC/MD simulations
In the final example, GCMC/MD simulations were used to produced larger RMSD values (approximately 1.0 Å) despite
hydrate the buried binding pocket of bovine pancreatic trypsin achieving full hydration of the binding pocket significantly
inhibitor (BPTI), and the most probable water positions were faster than MD simulations. The larger fluctuations observed in
compared with crystallographic data.68 The binding pocket was the random coil may be due to the assignment of random
defined as the region within 4.2 Å of the geometric center of velocities (from a Maxwell−Boltzmann distribution48,49) to
the Cα atoms of the Tyr10 and Asn43 residues.26 The average newly inserted water molecules. If maintaining correct protein
occupancy of each water site and the total number of observed dynamics is important, this issue could be mitigated by
water molecules within binding pocket are shown in Figure 5, increasing the length of the MD component of each cycle,
while data for each replica are presented in Figures S7−S10. reducing the number of attempted water molecule transfers
GCMC/MD simulations, initiated with an empty binding during the simulation.
pocket, reached equilibrium (defined as an occupancy of
≥0.75) within 0.3 ns, while MD simulations took approx-
imately 2.7 ns to fill all three water sites. Running on four cores
■ CONCLUSION
py-MCMD allows users to alter most of the parameters
of an Intel E5-2630v4 CPU and an NVIDIA RTX 2080TI governing a simulation’s behavior, and the strategy of linking
GPU, the GCMC/MD simulations were 1.8 times (total CPU two existing simulation engines, GOMC and NAMD, with an
+ GPU computational time) slower than MD simulations. external Python script provides tremendous flexibility. With
Considering the significant reduction in the time scale required this approach, MD configurational sampling can be incorpo-
to fully hydrate the binding pocket, the GCMC/MD rated into Monte Carlo simulations in any ensemble supported
simulations produced a computational efficiency approximately by GOMC. Additionally, Monte Carlo moves may be
5 times better than that of the MD simulations. integrated into any equilibrium NVT or NPT MD simulation.
Average water positions within the binding pocket were It is also possible to develop new sampling strategies with
calculated to be within 0.38 Å of experiment,68 as shown in minor revisions to py-MCMD, using a combination of MC to
Figure 5. The agreement with experiment is slightly better in propose trial configurations and nonequilibrium MD for
this work than in Samways et al.,26 whose results were within relaxation of the local structure.23,94
0.6 Å of experiment. The difference may be due to the choice Predictions of the hybrid MC/MD simulations of bulk SPC/
of force field: in the study by Samways et al., calculations were E water performed with py-MCMD were in close agreement
performed with the AMBER ff14sb force field91 for BPTI and with reference MC simulations. Substantial gains in computa-
the original TIP3P force field92 for water, while in this work, tional efficiency were observed for the MC/MD approach
CHARMM36 was used for BPTI and mTIP3P74 for water. compared with MC, ranging from a 2-fold improvement for
To understand the dynamics of hydration of the BPTI GCMC/MD simulations at 510 K to a 136-fold improvement
binding pocket, fluctuations of the random coil between the α- for NPT simulations at 298 K. NVT MC/MD simulations with
helix and the β-sheet (residues 36−47) were calculated with an intrabox swap move were used to rapidly equilibrate the
respect to the crystallographic structure for a system with a density of water in a system containing a graphene slit pore.
dehydrated binding pocket. The backbone of residues 36−47 The coupled−decoupled configurational-bias Monte Carlo
was clustered using the quality threshold (QT) algorithm93 as (CD−CBMC) algorithm37 substantially improved the sam-
implemented in VMD.73 Figure 6 shows the top three random pling efficiency for the transfer of water molecules. MC/MD
coil clusters for the first 2 ns of one MD simulation. The simulations using CD−CBMC required 25 times fewer cycles
random coil was collapsed for the majority of the first 2 ns of to reach equilibrium than MC/MD simulations using an
the MD simulation (green lines in Figure 6A), which blocked unbiased single random trial position, with a modest 25%
water from entering Site 1 of the binding pocket. Site 1 became increase in computational cost, and 66 times fewer cycles than
accessible only when a random fluctuation of the coil resulted prior hybrid simulations that used the cavity-bias method.19
in opening of the pocket. On the other hand, the GCMC/MD Finally, GCMC/MD simulations used to hydrate a buried
approach rapidly hydrated the binding pocket, especially Site 1, binding pocket in BPTI demonstrated a 5-fold improvement in
preventing the collapse of the binding pocket and enabling computational efficiency compared with MD simulations, and
further hydration through diffusion of surrounding waters in the water molecule locations observed in the GCMC/MD
addition to molecule transfers via GCMC/MD. An example of simulations were within 0.38 Å of the crystallographic water
random coil behavior for the dehydrated binding pocket in the sites. However, subtle differences were observed in protein
MD simulation is provided in Movie S1. dynamics in MD and GCMC/MD simulations. The GCMC/
Subtle differences were observed between the MD and MD simulations exhibited larger fluctuations of the random
GCMC/MD simulations in the stability of the fully hydrated coil (residues 36−47) than the MD simulations, which we
BPTI binding pocket. Figure 6 shows the root-mean-square attribute to velocity initialization of inserted water in the
deviation (RMSD) values of residues 36−47 with respect to GCMC/MD simulations to random values drawn from a
the crystallographic structure for one trajectory generated from Maxwell−Boltzmann distribution. It is expected that these
MD or GCMC/MD simulations after the backbone of residues differences could be addressed by increasing the length of the
in α-helices and β-sheets (residues 3−6, 18−24, 29−35, and MD component of each cycle in the GCMC/MD simulation.
48−55) were aligned with the crystallographic structure. Plots As an MC engine, GOMC includes a number of advanced
showing fluctuations of the random coil for each of the 10 configurational-bias algorithms that support the insertion of
replicas of MD and GCMC/MD simulations are provided in molecules having more complex topologies with acceptance
Figure S11. During the first 2 ns of the MD simulation, the rates that are 1−2 orders of magnitude greater than those of
4990 https://doi.org/10.1021/acs.jctc.1c00911
J. Chem. Theory Comput. 2022, 18, 4983−4994
Journal of Chemical Theory and Computation pubs.acs.org/JCTC Article

naive approaches.37−40 While the examples provided here

focused on water, it is straightforward to perform similar
■ ACKNOWLEDGMENTS
This research was partially supported by the National Institutes
calculations with larger, more complex molecules40 without of Health (Grant P41-GM104601) and the National Science
modification of the software. In addition to the applications Foundation (Grants OAC-1642406 and OAC-1835713).
provided here, py-MCMD could be used to simulate other Some of the computational resources used in this work were
phenomena, such as diffusion16 or gas adsorption in provided by the Wayne State University Grid computing
polymers.90 environment.

■ ASSOCIATED CONTENT
* Supporting Information
sı
■ REFERENCES
(1) Phillips, J. C.; Hardy, D. J.; Maia, J. D. C.; Stone, J. E.; Ribeiro, J.
The Supporting Information is available free of charge at V.; Bernardi, R. C.; Buch, R.; Fiorin, G.; Hénin, J.; Jiang, W.;
https://pubs.acs.org/doi/10.1021/acs.jctc.1c00911. McGreevy, R.; Melo, M. C. R.; Radak, B.; Skeel, R. D.; Singharoy, A.;
Wang, Y.; Roux, B.; Aksimentiev, A.; Luthey-Schulten, Z.; Kalé, L. V.;
Simulation parameters and results and summary of Schulten, K.; Chipot, C.; Tajkhorshid, E. Scalable molecular dynamics
Monte Carlo moves, (PDF) on CPU and GPU architectures with NAMD. J. Chem. Phys. 2020,
153, 044130.
Movie showing results for BPTI (MOV) (2) Enkavi, G.; Javanainen, M.; Kulig, W.; Róg, T.; Vattulainen, I.
Multiscale Simulations of Biological Membranes: The Challenge To

■ AUTHOR INFORMATION
Corresponding Authors
Understand Biological Phenomena in a Living Substance. Chem. Rev.
2019, 119, 5607−5774.
(3) Bochicchio, A.; Brandner, A. F.; Engberg, O.; Huster, D.;
Böckmann, R. A. Spontaneous membrane nanodomain formation in
Jeffrey Potoff − Department of Chemical Engineering and the absence or presence of the neurotransmitter serotonin. Front. Cell
Materials Science, Wayne State University, Detroit, Michigan Dev. Biol. 2020, 8, 601145.
48202, United States; orcid.org/0000-0002-4421-8787; (4) Durrant, J. D.; McCammon, J. A. Molecular dynamics
Phone: +1-313-577-9357; Email: [email protected] simulations and drug discovery. BMC Biol. 2011, 9, No. 71.
Emad Tajkhorshid − Theoretical and Computational (5) De Vivo, M.; Masetti, M.; Bottegoni, G.; Cavalli, A. Role of
Biophysics Group, NIH Center for Macromolecular Modeling molecular dynamics and related methods in drug discovery. J. Med.
and Bioinformatics, Beckman Institute for Advanced Science Chem. 2016, 59, 4035−4061.
and Technology, University of Illinois at (6) Wang, L.; Wu, Y.; Deng, Y.; Kim, B.; Pierce, L.; Krilov, G.;
Urbana−Champaign, Urbana, Illinois 61801, United States; Lupyan, D.; Robinson, S.; Dahlgren, M. K.; Greenwood, J.; Romero,
orcid.org/0000-0001-8434-1010; Phone: +1-217-244- D. L.; Masse, C.; Knight, J. L.; Steinbrecher, T.; Beuming, T.; Damm,
W.; Harder, E.; Sherman, W.; Brewer, M.; Wester, R.; Murcko, M.;
6914; Email: [email protected] Frye, L.; Farid, R.; Lin, T.; Mobley, D. L.; Jorgensen, W. L.; Berne, B.
Authors J.; Friesner, R. A.; Abel, R. Accurate and Reliable Prediction of
Relative Ligand Binding Potency in Prospective Drug Discovery by
Mohammad Soroush Barhaghi − Theoretical and Way of a Modern Free-Energy Calculation Protocol and Force Field.
Computational Biophysics Group, NIH Center for J. Am. Chem. Soc. 2015, 137, 2695−2703.
Macromolecular Modeling and Bioinformatics, Beckman (7) Klauda, J. B.; Venable, R. M.; Freites, J. A.; O’Connor, J. W.;
Institute for Advanced Science and Technology, University of Tobias, D. J.; Mondragon-Ramirez, C.; Vorobyov, I.; MacKerell, A.
Illinois at Urbana−Champaign, Urbana, Illinois 61801, D., Jr.; Pastor, R. W. Update of the CHARMM All-Atom Additive
United States Force Field for Lipids: Validation on Six Lipid Types. J. Phys. Chem. B
Brad Crawford − Department of Chemical Engineering and 2010, 114, 7830−7843.
Materials Science, Wayne State University, Detroit, Michigan (8) Venable, R. M.; Brown, F. L.; Pastor, R. W. Mechanical
properties of lipid bilayers from molecular dynamics simulation.
48202, United States; orcid.org/0000-0003-0638-7333
Chem. Phys. Lipids 2015, 192, 60−74.
Gregory Schwing − Department of Computer Science, Wayne (9) Muller, M. P.; Jiang, T.; Sun, C.; Lihan, M.; Pant, S.;
State University, Detroit, Michigan 48202, United States Mahinthichaichan, P.; Trifan, A.; Tajkhorshid, E. Characterization
David J. Hardy − Theoretical and Computational Biophysics of Lipid−Protein Interactions and Lipid-Mediated Modulation of
Group, NIH Center for Macromolecular Modeling and Membrane Protein Function through Molecular Simulations. Chem.
Bioinformatics, Beckman Institute for Advanced Science and Rev. 2019, 119, 6086−6161.
Technology, University of Illinois at Urbana−Champaign, (10) Jiang, T.; Wen, P.-C.; Trebesch, N.; Zhao, Z.; Pant, S.; Kapoor,
Urbana, Illinois 61801, United States; orcid.org/0000- K.; Shekhar, M.; Tajkhorshid, E. Computational Dissection of
0001-8533-1367 Membrane Transport at a Microscopic Level. Trends Biochem. Sci.
John E. Stone − Theoretical and Computational Biophysics 2020, 45, 202−216.
(11) Karplus, M.; Kuriyan, J. Molecular dynamics and protein
Group, NIH Center for Macromolecular Modeling and
function. Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 6679−6685.
Bioinformatics, Beckman Institute for Advanced Science and (12) Sharma, S.; Lindau, M. Molecular mechanism of fusion pore
Technology, University of Illinois at Urbana−Champaign, formation driven by the neuronal SNARE complex. Proc. Natl. Acad.
Urbana, Illinois 61801, United States Sci. U.S.A. 2018, 115, 12751−12756.
Loren Schwiebert − Department of Computer Science, Wayne (13) Bendahmane, M.; Bohannon, K. P.; Bradberry, M. M.; Rao, T.
State University, Detroit, Michigan 48202, United States C.; Schmidtke, M. W.; Abbineni, P. S.; Chon, N. L.; Tran, S.; Lin, H.;
Chapman, E. R.; Knight, J. D.; Anantharam, A. The synaptotagmin
Complete contact information is available at:
C2B domain calcium-binding loops modulate the rate of fusion pore
https://pubs.acs.org/10.1021/acs.jctc.1c00911 expansion. Mol. Biol. Cell 2018, 29, 834−845.
(14) Lee, E. H.; Hsin, J.; Sotomayor, M.; Comellas, G.; Schulten, K.
Notes Discovery Through the Computational Microscope. Structure 2009,
The authors declare no competing financial interest. 17, 1295−1306.

4991 https://doi.org/10.1021/acs.jctc.1c00911
J. Chem. Theory Comput. 2022, 18, 4983−4994
Journal of Chemical Theory and Computation pubs.acs.org/JCTC Article

(15) Dror, R. O.; Dirks, R. M.; Grossman, J.; Xu, H.; Shaw, D. E. A.; Swenson, D. W.; Henry, M.; Roet, S.; Silveira, A. choderalab/
Biomolecular Simulation: A Computational Microscope for Molecular openmmtools: 0.20.3 Bugfix Release, 2021. DOI: 10.5281/zeno-
Biology. Annu. Rev. Biophys. 2012, 41, 429−452. do.4639586.
(16) Heffelfinger, G. S.; van Swol, F. Diffusion in Lennard-Jones (35) Mezei, M. A cavity-biased (T, V, μ) Monte Carlo method for
fluids using dual control volume grand canonical molecular dynamics the computer simulation of fluids. Mol. Phys. 1980, 40, 901−906.
simulation (DCV-GCMD). J. Chem. Phys. 1994, 100, 7548−7552. (36) Crawford, B.; Potoff, J. py-MCMD: A Python Library for
(17) Woo, H. J.; Dinner, A. R.; Roux, B. Grand canonical Monte Performing Hybrid Monte Carlo/Molecular Dynamics Simulations with
Carlo simulations of water in protein environments. J. Chem. Phys. GOMC and NAMD. https://github.com/GOMC-WSU/py-MCMD
2004, 121, 6392−6400. (accessed January 2022).
(18) Henry, R. M.; Yu, C. H.; Rodinger, T.; Pomès, R. Functional (37) Martin, M.; Siepmann, J. Novel configurational-bias Monte
hydration and conformational gating of proton uptake in cytochrome Carlo method for branched molecules. Transferable potentials for
c oxidase. J. Mol. Biol. 2009, 387, 1165−1185. phase equilibria. 2. United-atom description of branched alkanes. J.
(19) Ben-Shalom, I. Y.; Lin, C.; Kurtzman, T.; Walker, R. C.; Gilson, Phys. Chem. B 1999, 103, 4508−4517.
M. K. Simulating water exchange to buried binding sites. J. Chem. (38) Soroush Barhaghi, M.; Torabi, K.; Nejahi, Y.; Schwiebert, L.;
Theory Comput. 2019, 15, 2684−2691. Potoff, J. J. Molecular exchange Monte Carlo: a generalized method
(20) Bodnarchuk, M. S.; Packer, M. J.; Haywood, A. Utilizing grand for identity exchanges in grand canonical Monte Carlo simulations. J.
canonical Monte Carlo methods in drug discovery. ACS Med. Chem. Chem. Phys. 2018, 149, 072318.
Lett. 2020, 11, 77−82. (39) Soroush Barhaghi, M.; Potoff, J. J. Prediction of phase equilibria
(21) Ross, G. A.; Russell, E.; Deng, Y.; Lu, C.; Harder, E. D.; Abel, and Gibbs free energies of transfer using molecular exchange Monte
R.; Wang, L. Enhancing water sampling in free energy calculations Carlo in the Gibbs ensemble. Fluid Phase Equilib. 2019, 486, 106−
with grand canonical Monte Carlo. J. Chem. Theory Comput. 2020, 16, 118.
6061−6076. (40) Nejahi, Y.; Barhaghi, M. S.; Schwing, G.; Schwiebert, L.; Potoff,
(22) De Joannis, J.; Coppock, P. S.; Yin, F.; Mori, M.; Zamorano, A.; J. Update 2.70 to “GOMC: GPU optimized Monte Carlo for the
Kindt, J. T. Atomistic simulation of cholesterol effects on miscibility simulation of phase equilibria and physical properties of complex
of saturated and unsaturated phospholipids: implications for liquid- fluids”. SoftwareX 2021, 13, 100627.
ordered/liquid-disordered phase coexistence. J. Am. Chem. Soc. 2011, (41) Panagiotopoulos, A. Z.; Quirke, N.; Stapleton, M.; Tildesley, D.
133, 3625−3634. J. Phase equilibria by simulation in the Gibbs ensemble alternative
(23) Fathizadeh, A.; Elber, R. A Mixed Alchemical and Equilibrium derivation, generalization and application to mixture and membrane
Dynamics to Simulate Heterogeneous Dense Fluids: Illustrations for equilibria. Mol. Phys. 1988, 63, 527−545.
Lennard-Jones Mixtures and Phospholipid Membranes. J. Chem. Phys. (42) Gartner, T. E.; Epps, T. H.; Jayaraman, A. Leveraging Gibbs
2018, 149, 072325. Ensemble Molecular Dynamics and Hybrid Monte Carlo/Molecular
(24) Deng, Y.; Roux, B. Computation of binding free energy with Dynamics for Efficient Study of Phase Equilibria. J. Chem. Theory
molecular dynamics and grand canonical Monte Carlo simulations. J. Comput. 2016, 12, 5501−5510.
Chem. Phys. 2008, 128, 115103. (43) Fass, J.; Sivak, D. A.; Crooks, G. E.; Beauchamp, K. A.;
(25) Pool, R.; Heringa, J.; Hoefling, M.; Schulz, R.; Smith, J. C.; Leimkuhler, B.; Chodera, J. D. Quantifying Configuration-Sampling
Feenstra, K. A. Enabling grand-canonical Monte Carlo: extending the Error in Langevin Simulations of Complex Molecular Systems.
flexibility of GROMACS through the grompy Python interface Entropy 2018, 20, No. 318.
module. J. Comput. Chem. 2012, 33, 1207−1214. (44) Brünger, A.; Brooks, C. B., III; Karplus, M. Stochastic
(26) Samways, M. L.; Bruce Macdonald, H. E.; Essex, J. W. grand: a boundary-conditions for molecular dynamics simulations of ST2
Python module for grand canonical water sampling in OpenMM. J. water. Chem. Phys. Lett. 1984, 105, 495−500.
Chem. Inf. Model. 2020, 60, 4436−4441. (45) Leimkuhler, B.; Matthews, C. Robust and efficient configura-
(27) Wahl, J.; Smieško, M. Assessing the predictive power of relative tional molecular sampling via Langevin dynamics. J. Chem. Phys. 2013,
binding free energy calculations for test cases involving displacement 138, No. 174102.
of binding site water molecules. J. Chem. Inf. Model. 2019, 59, 754− (46) Leimkuhler, B.; Matthews, C. Efficient molecular dynamics
765. using geodesic integration and solvent−solute splitting. Proc. R. Soc. A
(28) Ben-Shalom, I. Y.; Lin, Z.; Radak, B. K.; Lin, C.; Sherman, W.; 2016, 472, No. 20160138.
Gilson, M. K. Accounting for the central role of interfacial water in (47) Leimkuhler, B.; Matthews, C. Rational Construction of
protein-ligand binding free energy calculations. J. Chem. Theory Stochastic Numerical Methods for Molecular Sampling. Appl. Math.
Comput. 2020, 16, 7883−7894. Res. eXpress 2012, 2013, 34−56.
(29) Im, W.; Seefeld, S.; Roux, B. A Grand Canonical Monte Carlo- (48) Boltzmann, L. Weitere studien über das Wärmegleichgewicht
Brownian Dynamics Algorithm for Simulating Ion Channels. Biophys. unter gasmolekülen. Sitzungsber. Kaiserl. Akad. Wiss., Math.-Naturwiss.
J. 2000, 79, 788−801. Cl. 1872, 66, 275−370.
(30) Solano, C. J.; Prajapati, J. D.; Pothula, K. R.; Kleinekathöfer, U. (49) Maxwell, M. A.; J, C. V. Illustrations of the dynamical theory of
Brownian dynamics approach including explicit atoms for studying gases.�Part I. On the motions and collisions of perfectly elastic
ion permeation and substrate translocation across nanopores. J. Chem. spheres. London Edinburgh Philos. Mag. J. Sci. 1860, 19, 19−32.
Theory Comput. 2018, 14, 6701−6713. (50) Allen, M. P.; Tildesley, D. J. Computer Simulation of Liquids;
(31) Hu, J.; Ma, A.; Dinner, A. R. Monte Carlo simulations of Oxford University Press: New York, 1987.
biomolecules: the MC module in CHARMM. J. Comput. Chem. 2006, (51) Ewald, P. P. Die Berechnung optischer und elektrostatischer
27, 203−216. Gitterpotentiale. Ann. Phys. 1921, 369, 253−287.
(32) Plimpton, S. J. Fast parallel algorithms for short-range (52) Darden, T.; York, D.; Pedersen, L. Particle Mesh Ewald: an N·
molecular dynamics. J. Comput. Phys. 1995, 117, 1−19. log(N) Method for Ewald Sums in Large Systems. J. Chem. Phys.
(33) Eastman, P.; Swails, J.; Chodera, J. D.; McGibbon, R. T.; Zhao, 1993, 98, 10089−10092.
Y.; Beauchamp, K. A.; Wang, L.-P.; Simmonett, A. C.; Harrigan, M. (53) Essmann, U.; Perera, L.; Berkowitz, M. L.; Darden, T.; Lee, H.;
P.; Stern, C. D.; Wiewiora, R. P.; Brooks, B. R.; Pande, V. S. Pedersen, L. G. A smooth particle mesh Ewald method. J. Chem. Phys.
OpenMM 7: Rapid development of high performance algorithms for 1995, 103, 8577−8593.
molecular dynamics. PLoS Comput. Biol. 2017, 13, No. e1005659. (54) Bai, P.; Siepmann, J. I. Assessment and optimization of
(34) Chodera, J.; Rizzi, A.; Naden, L.; Beauchamp, K.; Grinaway, P.; configurational-bias Monte Carlo particle swap strategies for
Fass, J.; Wade, A.; Rustenburg, B.; Ross, G. A.; Krämer, A.; simulations of water in the Gibbs ensemble. J. Chem. Theory Comput.
Macdonald, H. B.; Rodríguez-Guerra, J.; dominicrufa; Simmonett, 2017, 13, 431−440.

4992 https://doi.org/10.1021/acs.jctc.1c00911
J. Chem. Theory Comput. 2022, 18, 4983−4994
Journal of Chemical Theory and Computation pubs.acs.org/JCTC Article

(55) Shah, J. K.; Maginn, E. J. A general and efficient Monte Carlo (73) Humphrey, W.; Dalke, A.; Schulten, K. VMD − Visual
method for sampling intramolecular degrees of freedom of branched Molecular Dynamics. J. Mol. Graphics 1996, 14, 33−38.
and cyclic molecules. J. Chem. Phys. 2011, 135, 134121. (74) MacKerell, A. D., Jr.; Bashford, D.; Bellott, M.; Dunbrack, R. L.,
(56) Miyamoto, S.; Kollman, P. A. Settle: an analytical version of the Jr.; Evanseck, J. D.; Field, M. J.; Fischer, S.; Gao, J.; Guo, H.; Ha, S.;
SHAKE and RATTLE algorithm for rigid water molecules. J. Comput. Joseph-McCarthy, D.; Kuchnir, L.; Kuczera, K.; Lau, F. T. K.; Mattos,
Chem. 1992, 13, 952−962. C.; Michnick, S.; Ngo, T.; Nguyen, D. T.; Prodhom, B.; Reiher, I. W.
(57) Schneider, T.; Stoll, E. Molecular-dynamics study of a three- E.; Roux, B.; Schlenkrich, M.; Smith, J.; Stote, R.; Straub, J.;
dimensional one-component model for distortive phase transitions. Watanabe, M.; Wiorkiewicz-Kuczera, J.; Yin, D.; Karplus, M. All-atom
Phys. Rev. B 1978, 17, 1302−1322. empirical potential for molecular modeling and dynamics studies of
(58) Tuckerman, M.; Berne, B. J.; Martyna, G. J. Reversible multiple proteins. J. Phys. Chem. B 1998, 102, 3586−3616.
time scale molecular dynamics. J. Chem. Phys. 1992, 97, 1990−2001. (75) Brooks, B. R.; Brooks, C. L.; Mackerell, A. D.; Nilsson, L.;
(59) Berendsen, H. J. C.; Grigera, J. R.; Straatsma, T. P. The Missing Petrella, R. J.; Roux, B.; Won, Y.; Archontis, G.; Bartels, C.; Boresch,
Term in Effective Pair Potentials. J. Phys. Chem. 1987, 91, 6269− S.; Caflisch, A.; Caves, L.; Cui, Q.; Dinner, A. R.; Feig, M.; Fischer, S.;
6271. Gao, J.; Hodoscek, M.; Im, W.; Kuczera, K.; Lazaridis, T.; Ma, J.;
(60) Summers, A. Z.; Gilmer, J. B.; Iacovella, C. R.; Cummings, P. Ovchinnikov, V.; Paci, E.; Pastor, R. W.; Post, C. B.; Pu, J. Z.;
T.; McCabe, C. MoSDeF, a Python framework enabling large-scale Schaefer, M.; Tidor, B.; Venable, R. M.; Woodcock, H. L.; Wu, X.;
computational screening of soft matter: application to chemistry- Yang, W.; York, D. M.; Karplus, M. CHARMM: The biomolecular
property relationships in lubricating monolayer films. J. Chem. Theory simulation program. J. Comput. Chem. 2009, 30, 1545−1614.
Comput. 2020, 16, 1779−1793. (76) Jo, S.; Kim, T.; Iyer, V. G.; Im, W. CHARMM-GUI: a Web-
(61) Klein, C.; Summers, A. Z.; Thompson, M. W.; Gilmer, J. B.; based Graphical User Interface for CHARMM. J. Comput. Chem.
McCabe, C.; Cummings, P. T.; Sallai, J.; Iacovella, C. R. Formalizing 2008, 29, 1859−1865.
atom-typing and the dissemination of force fields with foyer. Comput. (77) Brooks, B. R.; Bruccoleri, R. E.; Olafson, B. D.; States, D. J.;
Mater. Sci. 2019, 167, 215−227. Swaminathan, S.; Karplus, M. CHARMM: A Program for Macro-
(62) Thompson, M. W.; Gilmer, J. B.; Matsumoto, R. A.; Quach, C.
molecular Energy, Minimization, and Dynamics Calculations. J.
D.; Shamaprasad, P.; Yang, A. H.; Iacovella, C. R.; McCabe, C.;
Comput. Chem. 1983, 4, 187−217.
Cummings, P. T. Towards molecular simulations that are transparent,
(78) Best, R. B.; Zhu, X.; Shim, J.; Lopes, P. E. M.; Mittal, J.; Feig,
reproducible, usable by others, and extensible (TRUE). Mol. Phys.
M.; MacKerell, A. D. Optimization of the additive CHARMM all-
2020, 118, No. e1742938.
atom protein force field targeting improved sampling of the backbone
(63) Matsumoto, R.; Thompson, M.; DeFever, R. Pore-builder.
ϕ, ψ and side-chain χ1 and χ2 dihedral angles. J. Chem. Theory Comput.
https://github.com/rmatsum836/Pore-Builder (accessed March
2012, 8, 3257−3273.
2021).
(79) Lee, S.; Tran, A.; Allsopp, M.; Lim, J. B.; Hénin, J.; Klauda, J. B.
(64) Nejahi, Y.; Soroush Barhaghi, M.; Mick, J.; Jackman, B.;
Rushaidat, K.; Li, Y.; Schwiebert, L.; Potoff, J. GOMC: GPU CHARMM36 United Atom Chain Model for Lipids and Surfactants.
optimized Monte Carlo for the simulation of phase equilibria and J. Phys. Chem. B 2014, 118, 547−556.
physical properties of complex fluids. SoftwareX 2019, 9, 20−27. (80) Huang, J.; MacKerell, A. D. CHARMM36 all-atom additive
(65) Shirts, M. R.; Chodera, J. D. Statistically optimal analysis of protein force field: Validation based on comparison to NMR data. J.
samples from multiple equilibrium states. J. Chem. Phys. 2008, 129, Comput. Chem. 2013, 34, 2135−2145.
124105. (81) Lee, J.; Cheng, X.; Swails, J. M.; Yeom, M.; Eastman, P. K.;
(66) Chodera, J. D.; Swope, W. C.; Pitera, J. W.; Seok, C.; Dill, K. A. Lemkul, J. A.; Wei, S.; Buckner, J.; Jeong, J.; Qi, Y.; Jo, S.; Pande, V.
Use of the Weighted Histogram Analysis Method for the Analysis of S.; Case, D. A.; Brooks, C. L., III; MacKerell, A. D., Jr.; Klauda, J. B.;
Simulated and Parallel Tempering Simulations. J. Chem. Theory Im, W. CHARMM-GUI Input Generator for NAMD, GROMACS,
Comput. 2007, 3, 26−41. AMBER, OpenMM, and CHARMM/OpenMM Simulations Using
(67) Striolo, A.; Chialvo, A. A.; Cummings, P. T.; Gubbins, K. E. the CHARMM36 Additive Force Field. J. Chem. Theory Comput.
Water adsorption in carbon-slit nanopores. Langmuir 2003, 19, 2016, 12, 405−413.
8583−8591. (82) Huang, J.; Rauscher, S.; Nawrocki, G.; Ran, T.; Feig, M.; de
(68) Wlodawer, A.; Walter, J.; Huber, R.; Sjölin, L. Structure of Groot, B. L.; Grubmüller, H.; MacKerell, A. D., Jr CHARMM36m: an
bovine pancreatic trypsin inhibitor: results of joint neutron and X-ray improved force field for folded and intrinsically disordered proteins.
refinement of crystal form II. J. Mol. Biol. 1984, 180, 301−329. Nat. Methods 2017, 14, 71−73.
(69) Burley, S. K.; Bhikadiya, C.; Bi, C.; Bittrich, S.; Chen, L.; (83) Boonstra, S.; Onck, P. R.; van der Giessen, E. CHARMM
Crichlow, G. V.; Christie, C. H.; Dalenberg, K.; Di Costanzo, L.; TIP3P water model suppresses peptide folding by solvating the
Duarte, J. M.; Dutta, S.; Feng, Z.; Ganesan, S.; Goodsell, D. S.; Ghosh, unfolded state. J. Phys. Chem. B 2016, 120, 3692−3698.
S.; Green, R. K.; Guranović, V.; Guzenko, D.; Hudson, B. P.; Lawson, (84) Feller, S. E.; Zhang, Y.; Pastor, R. W.; Brooks, B. R. Constant
C. L.; Liang, Y.; Lowe, R.; Namkoong, H.; Peisach, E.; Persikova, I.; pressure molecular dynamics simulation: The Langevin piston
Randle, C.; Rose, A.; Rose, Y.; Sali, A.; Segura, J.; Sekharan, M.; Shao, method. J. Chem. Phys. 1995, 103, 4613−4621.
C.; Tao, Y.-P.; Voigt, M.; Westbrook, J. D.; Young, J. Y.; Zardecki, C.; (85) Hoover, W. G. Canonical Dynamics: Equilibrium Phase-Space
Zhuravleva, M. RCSB Protein Data Bank: powerful new tools for Distributions. Phys. Rev. A 1985, 31, 1695−1697.
exploring 3D structures of biological macromolecules for basic and (86) Hoover, W. G. Constant-Pressure Equations of Motion. Phys.
applied research and education in fundamental biology, biomedicine, Rev. A 1986, 34, 2499−2500.
biotechnology, bioengineering and energy sciences. Nucleic Acids Res. (87) Ong, E. E.; Liow, J.-L. The temperature-dependent structure,
2021, 49, D437−D451. hydrogen bonding and other related dynamic properties of the
(70) Berman, H. M.; Westbrook, J.; Feng, Z.; Gilliland, G.; Bhat, T. standard TIP3P and CHARMM-modified TIP3P water models. Fluid
N.; Weissig, H.; Shindyalov, I. N.; Bourne, P. E. The Protein Data Phase Equilib. 2019, 481, 55−65.
Bank. Nucleic Acids Res. 2000, 28, 235−242. (88) Sakamaki, R.; Sum, A. K.; Narumi, T.; Yasuoka, K. Molecular
(71) Berman, H.; Henrick, K.; Nakamura, H. Announcing the dynamics simulations of vapor/liquid coexistence using the non-
worldwide Protein Data Bank. Nat. Struct. Biol. 2003, 10, 980. polarizable water models. J. Chem. Phys. 2011, 134, 124708.
(72) Ribeiro, J. V.; Bernardi, R. C.; Rudack, T.; Stone, J. E.; Phillips, (89) Boulougouris, G. C.; Economou, I. G.; Theodorou, D. N.
J. C.; Freddolino, P. L.; Schulten, K. QwikMD-Integrative Molecular Engineering a molecular model for water phase equilibrium over a
Dynamics Toolkit for Novices and Experts. Sci. Rep. 2016, 6, 26536. wide temperature range. J. Phys. Chem. B 1998, 102, 1029−1035.

4993 https://doi.org/10.1021/acs.jctc.1c00911
J. Chem. Theory Comput. 2022, 18, 4983−4994
Journal of Chemical Theory and Computation pubs.acs.org/JCTC Article

(90) Gartner, T. E.; Jayaraman, A. Macromolecular ‘size’ and

‘hardness’ drives structure in solvent-swollen blends of linear, cyclic,
and star polymers. Soft Matter 2018, 14, 411−423.
(91) Maier, J. A.; Martinez, C.; Kasavajhala, K.; Wickstrom, L.;
Hauser, K. E.; Simmerling, C. ff14SB: Improving the Accuracy of
Protein Side Chain and Backbone Parameters from ff99SB. J. Chem.
Theory Comput. 2015, 11, 3696−3713.
(92) Jorgensen, W.; Chandrasekhar, J.; Madura, J. D.; Impey, R. W.;
Klein, M. L. Comparison of Simple Potential Functions for Simulating
Liquid Water. J. Chem. Phys. 1983, 79, 926−935.
(93) Heyer, L. J.; Kruglyak, S.; Yooseph, S. Exploring expression
data: identification and analysis of coexpressed genes. Gen. Res. 1999,
9, 1106−1115.
(94) Chen, Y.; Roux, B. Generalized Metropolis Acceptance
Criterion for Hybrid Non-equilibrium Molecular Dynamics−Monte
Carlo Simulations. J. Chem. Phys. 2015, 142, 024101.

■ NOTE ADDED AFTER ASAP PUBLICATION

This paper as published ASAP on May 27, 2022, with scientific
errors in the Supporting Information. The corrected version
was reposted on June 3, 2022.

Recommended by ACS
Evidence That Less Can Be More for Transferable Force
Fields
Bumjoon Seo and Brett M. Savoie
FEBRUARY 06, 2023
JOURNAL OF CHEMICAL INFORMATION AND MODELING READ

Flow-Matching: Efficient Coarse-Graining of Molecular

Dynamics without Forces
Jonas Köhler, Frank Noé, et al.
JANUARY 20, 2023
JOURNAL OF CHEMICAL THEORY AND COMPUTATION READ

Re-Balancing Replica Exchange with Solute Tempering for

Sampling Dynamic Protein Conformations
Yumeng Zhang, Jianhan Chen, et al.
FEBRUARY 15, 2023
JOURNAL OF CHEMICAL THEORY AND COMPUTATION READ

Statistical Mechanical Design Principles for Coarse-Grained

Interactions across Different Conformational Free Energy
Surfaces
Jaehyeok Jin and Gregory A. Voth
FEBRUARY 02, 2023
THE JOURNAL OF PHYSICAL CHEMISTRY LETTERS READ

Get More Suggestions >

4994 https://doi.org/10.1021/acs.jctc.1c00911
J. Chem. Theory Comput. 2022, 18, 4983−4994