Molecular Dynamics Simulations

in GROMACS

presented by
Jan Schulze
Outline

1. Basic principles of MD simulations

2. MD simulations in GROMACS
▪ Main steps in performing MD simulations

▪ Important computational aspects Focus

▪ Different types of input files and what they mean (technology)
▪ Different types of output files from simulation

▪ Some colorful visualizations

Remark: Case study not discussed today.

▪ Same procedure, several times in a row
▪ Today: Focus on GROMACS technology behind MD simulations

Molecular Dynamics Simulations in GROMACS 11.12.2017 Jan Schulze 2

Basic principles of MD simulations

Newton’s equations of motion for every atom:

𝑑 𝑑
𝐫 = 𝐮i and 𝑚𝑖 ⋅ 𝐮 = 𝐟i
𝑑𝑡 i 𝑑𝑡 i

The force fi is the gradient of the potential V :

𝜕
𝐟i = − 𝑉 𝐫N
𝜕𝐫i
Contributions to the potential V :
▪ Intermolecular interactions
 Lennard-Jones potential (repulsion, van-der-Waals) Force field
 Coulomb potential (charges)
▪ Functional terms (models)
▪ Intramolecular interactions This is how we get
 Stretching vibrations molecules! ▪ Model parameters
 Bending vibrations
 Internal rotation
e.g. Optimized Potentials for
Liquid Systems force field

Molecular Dynamics Simulations in GROMACS 11.12.2017 Jan Schulze 3

Computational aspects (1)

Parallelization: Equations of motion for every atom can be solved independently

from one time step to the next.
▪ But: Need all positions together to calculate forces for next step.

Δ𝐫1 Δ𝐫1
Δ𝐫2 V(r1,…,rN) Δ𝐫2 V(r1,…,rN)

t1 Δ𝐫3 t2 Δ𝐫3 t3
Δ𝐫𝑁 Δ𝐫𝑁

Periodic boundary
conditions: Avoid simulation of wall effects in bulk systems.
▪ Allow atom to “pass” the wall, but relocate it to the opposite wall.
▪ Atoms can interact through periodic walls as well.
▪ Leads to bulk-like behavior, even at the walls.

Molecular Dynamics Simulations in GROMACS 11.12.2017 Jan Schulze 4

Computational aspects (2)

Controls: Simulations of practical relevance:

▪ NVT simulation (closed system, heat transfer)
▪ NpT simulation (closed system, heat and work transfer)

Control structures to keep T, p constant:

Barostat: Rescales atom positions (box volume) to
keep pressure constant (adjust potential
energy).
Thermostat: Rescales atom velocities to keep temperature
constant (adjust kinetic energy).

Equilibration: Usually, simulations start in non-equilibrium state. Molecular

positions and velocities do not correspond with equilibrium.

▪ First, perform equilibration simulation until thermodynamic equilibrium

reached.
▪ Then extend simulation time to simulate actual equilibrium state and apply
time-averaging for analysis.

Molecular Dynamics Simulations in GROMACS 11.12.2017 Jan Schulze 5

Computational aspects (3)

Time and
length scales: Interactions on a molecular level are very fast and
occur on a very small length scale
▪ Characteristic length scale: Nanometers
▪ Characteristic time scale: Femtoseconds (10 - 6 ns)

Number of molecules: Affects computational effort but also the quality of data
▪ Too small: High statistical influence in simulation results,
low-quality data, results loose on generality
▪ Too high: Simulation takes “forever”

Typical number for pure component: ~ 1000 molecules

Need refinement study to justify choice.

Molecular Dynamics Simulations in GROMACS 11.12.2017 Jan Schulze 6

Main steps in a GROMACS simulation

1. Define molecular
structure of components

2. Set up force field,

make topology

3. Make box(es) and 5. Perform an energy

insert molecules minimization

4. Merge all necessary 6. Run the simulation

files

7. Check for equilibration,

analyze results

Molecular Dynamics Simulations in GROMACS 11.12.2017 Jan Schulze 7

File structure and information flow in GROMACS
analysis.xtc
▪ Keeping track of the “entangled” file structure is
probably one of the biggest issues in performing
Analysis
GROMACS simulations.
▪ Systematic procedure from previous slide helps

▪ For every simulation we need three types of files:

main_simulation.gro main_simulation.trr more files ...
 Topology, force field (.top)
 Simulation settings (.mdp)
Simulation  Initial box, atom coordinates (.gro)

main_simulation.tpr minimization.gro minimization.trr more files ...

Energy minimization
simulation

minimization.tpr

sim_settings.mdp topology.top minim_settings.mdp box_file.gro

oplsaa.ff/forcefield.itp MEG.itp
MEG.pdb

ffnonbonded.itp ffnonbonded.itp reference to file

combination of files or
information flow

Molecular Dynamics Simulations in GROMACS 11.12.2017 Jan Schulze 8

File structure and information flow in GROMACS
analysis.xtc
▪ Keeping track of the “entangled” file structure is
probably one of the biggest issues in performing
Analysis
GROMACS simulations.
▪ Systematic procedure from previous slide helps

▪ For every simulation we need three types of files:

main_simulation.gro main_simulation.trr more files ...
 Topology, force field (.top)
 Simulation settings (.mdp)
Simulation  Initial box, atom coordinates (.gro)

main_simulation.tpr minimization.gro minimization.trr more files ...

Energy minimization
simulation

minimization.tpr

sim_settings.mdp topology.top minim_settings.mdp box_file.gro

oplsaa.ff/forcefield.itp MEG.itp
MEG.pdb

ffnonbonded.itp ffnonbonded.itp reference to file

combination of files or
information flow

Molecular Dynamics Simulations in GROMACS 11.12.2017 Jan Schulze 9

Step 1 – Define molecular structure of components
COMPND UNNAMED
1. DefineGENERATED
AUTHOR molecularBY OPEN BABEL 2.3.90 Geometry file
structure
HETATM of1 components
C MEG 1 1.263 -0.256 0.128 1.00 0.00 C (.pdb)
HETATM 2 C MEG 1 0.367 -0.637 1.303 1.00 0.00 C
HETATM 3 O MEG 1 -0.640 0.357 1.491 1.00 0.00 O
HETATM 4 O MEG 1 2.270 0.659 0.558 1.00 0.00 O
HETATM 5 H MEG 1 0.684 0.221 -0.670 1.00 0.00 H
HETATM 6 H MEG 1 1.771 -1.141 -0.264 1.00 0.00 H
HETATM 7 H MEG 1 0.946 -0.718 2.228 1.00 0.00 H
HETATM 8 H MEG 1 -0.141 -1.583 1.100 1.00 0.00 H
HETATM 9 H MEG 1 -1.088 0.148 2.327 1.00 0.00 H
HETATM 10 H MEG 1 2.718 0.981 -0.242 1.00 0.00 H
CONECT 1 2 4 5 6 atom positions in the molecule
CONECT 2 1 3 7 8
CONECT 3 2 9
CONECT 4 1 10
CONECT 5 1 connectivity
CONECT 6 1
CONECT 7 2
CONECT 8 2
CONECT 9 3
CONECT 10 4
MASTER 0 0 0 0 0 0 0 0 10 0 10 0
END

MEG.pdb

▪ Simple file structure that contains a lot of

important information

Molecular Dynamics Simulations in GROMACS 11.12.2017 Jan Schulze 10

File structure and information flow in GROMACS
analysis.xtc

reference to file
combination of files or
Analysis information flow

main_simulation.gro main_simulation.trr more files ...

Simulation

main_simulation.tpr minimization.gro minimization.trr more files ...

Energy minimization
simulation

minimization.tpr

sim_settings.mdp topology.top minim_settings.mdp box_file.gro

oplsaa.ff/forcefield.itp MEG.itp
MEG.pdb

ffnonbonded.itp ffnonbonded.itp
Step 2 – Topology & force field Step 1 – Define molecular structure

Molecular Dynamics Simulations in GROMACS 11.12.2017 Jan Schulze 11

Step 2 – Make topology
topology.top
1. Define molecular
structure of components

oplsaa.ff/forcefield.itp Component_1.itp

2. Set up force field,

make topology ffnonbonded.itp
ffnonbonded.itp

What is the topology?

▪ A set of all constant attributes of our system ; Force field
#include "oplsaa.ff/forcefield.itp"
 Information about force field #include „Component_1.itp“
 Number of molecules
[ system ]
; Name
How does it look like? Liquid Simulation

▪ Information about force field is structured into [ molecules ]

; Compound #mols
multiple sub-files MEG 1000

▪ User has to set up the force field file(s) for

the component(s), e.g. Component_1.itp topology.top

Molecular Dynamics Simulations in GROMACS 11.12.2017 Jan Schulze 12

Step 2 – Set up force field

1. Define molecular
structure of components
The OPLS force field
Set of functional forms
2. Set up force field, ▪ Implementation of OPLS functional terms (and many more) is
make topology
already done in GROMACS library

Parameters
▪ Describe how atoms interact with other atoms
▪ User has to provide parameters that match with particular system
force field file (.itp)  Many parameter sets already implemented in GROMACS
that we have to create  User (only) has to refer to correct parameter set for every atom
(one for every species)
 This is realized as specification of an atom type for every atom

Functional forms
of OPLS force field:

Molecular Dynamics Simulations in GROMACS 11.12.2017 Jan Schulze 13

Step 2 – Set up force field

1. Define molecular
structure of components Tasks in step 2
▪ State correct atom type for every atom in every molecule
2. Set up force field, ▪ Generate .itp force field file for every molecule
make topology ▪ Finally, generate .top topology for the system

Atom types? Why?  The same element in a different molecule

interacts in a different way

force field parameter force field parameter

(charge) of atom 1 (charge) of atom 2
𝑞1 ⋅ 𝑞2
Coulomb potential: 𝑉 𝑟12 = 𝑓 ⋅
𝑟12
Same element (Cl), but
different partial charges
Source: https://saylordotorg.github.io/text_general-chemistry-principles-patterns-and-applications-v1.0/s20-03-molecular-structure-and-acid-b.html

Molecular Dynamics Simulations in GROMACS 11.12.2017 Jan Schulze 14

How are atom types defined?

Ethylene Glycol

atom type code

atomtypes.atp

Molecular Dynamics Simulations in GROMACS 11.12.2017 Jan Schulze 15

Step 2 – Generate .itp file
[ moleculetype ]
; Name nrexcl
Use a script to generate .itp file
UNK
MEG 3 structure from geometry .pdb file
[ atoms ]
; nr type resnr residue atom cgnr charge mass
1 opls_800 1 UNK
MEG C00 1 -0.0329 12.0110
2 opls_801 1 UNK
MEG C01 1 -0.0330 12.0110
3 opls_802 1 UNK
MEG O02 1 -0.5898 15.9990
4 opls_803 1 UNK
MEG O03 1 Very helpful – BUT atom types can be incorrect
-0.5898 15.9990
Browse…

5 opls_804 1 UNK
MEG H04 1 0.1066 1.0080
6 opls_805 1 UNK
MEG H05 1 (as well as bonds, angles or dihedrals)
0.1066 1.0080
7 opls_806 1 UNK
MEG H06 1 0.1065 1.0080
8 opls_807 1 UNK
MEG H07 1 0.1065 1.0080
9 opls_808 1 UNK
MEG H08 1 0.4097 1.0080
10 opls_809 1 UNK
MEG H09 1 0.4097 1.0080

[ bonds ]
; i j funct http://zarbi.chem.yale.edu/ligpargen/
2 1 1 0.1529 224262.400
3 2 1 0.1410 267776.000
4 1 1 0.1410 267776.000
5 1 1 0.1090 284512.000
6 1 1 0.1090 284512.000
7 2 1 0.1090 284512.000
8 2 1 0.1090 284512.000
9 3 1 0.0945 462750.400
10 4 1 0.0945 462750.400 Index sets for bonds, angles, torsion
[ angles ]
; i j k funct c0
1 2 3 1 109.500 418.400
2 1 4 1 109.500 418.400
2 1 5 1 110.700 313.800
2 1 6 1 110.700 313.800
1 2 7 1 110.700 313.800
1 2 8 1 110.700 313.800
2 3 9 1 108.500 460.240
1 4 10 1 108.500 460.240
Molecular
4 1 Dynamics
5 1Simulations
109.500 in GROMACS
292.880 MEG.itp
11.12.2017 Jan Schulze 16
5 1 6 1 107.800 276.144
Atom types

1. Define molecular
structure of components

2. Set up force field,

make topology

How to check and select correct atom types?

▪ The OPLS force field provides a list of all parametrized atom
types + description
▪ Read through this list and find atom types that apply on your
molecular structure
▪ Conduct literature research and see what other authors have
used

Molecular Dynamics Simulations in GROMACS 11.12.2017 Jan Schulze 17

File structure and information flow in GROMACS
analysis.xtc

reference to file
combination of files or
Analysis information flow

main_simulation.gro main_simulation.trr more files ...

Simulation

main_simulation.tpr minimization.gro minimization.trr more files ...

Energy minimization
simulation

minimization.tpr

Step 3 – Make a box

sim_settings.mdp topology.top
P minim_settings.mdp box_file.gro

oplsaa.ff/forcefield.itp P MEG.itp P MEG.pdb

P
ffnonbonded.itp
P ffnonbonded.itp
P

Molecular Dynamics Simulations in GROMACS 11.12.2017 Jan Schulze 18

MD skills – Generation of a liquid phase

▪ High density of liquid poses a problem

for insertion algorithms

▪ Atom shells overlap frequently, causing

the molecule insertion algorithm to fail

▪ How to choose the box dimensions then?

Box dimensions chosen according to realistic
density
Strategy for liquid box:
▪ Insert molecules into an oversized box
(factor ~ 10)
▪ Shrink the box dimensions in an NpT
simulation Analysis
▪ Simulation until equilibration at realistic
density reached (equilibration)

Molecular Dynamics Simulations in GROMACS 11.12.2017 Jan Schulze 19

File structure and information flow in GROMACS
analysis.xtc

reference to file
combination of files or
Analysis information flow

main_simulation.gro main_simulation.trr more files ...

Simulation

main_simulation.tpr minimization.gro minimization.trr more files ...

Energy minimization
simulation

minimization.tpr

sim_settings.mdp topology.top
P minim_settings.mdp box_file.gro
P
oplsaa.ff/forcefield.itp
P MEG.itp
P MEG.pdb
P
ffnonbonded.itp
P ffnonbonded.itp
P

Molecular Dynamics Simulations in GROMACS 11.12.2017 Jan Schulze 20

 http://manual.gromacs.org/documentation/5.1/user-guide/mdp-options.html

Simulation settings – Example: NpT simulation

...
o VARIOUS PREPROCESSING OPTIONS =
title = NPT
cpp = /lib/cpp
; OPTIONS FOR ELECTROSTATICS AND VDW = cutoff radius
; Method for doing electrostatics =
include = -I../top
coulombtype = PME
define =
rcoulomb = 1.0
; Method for doing Van der Waals =
; RUN CONTROL PARAMETERS =
integrator = md
MD integrator vdw-type = PME
rvdw = 1.0
; start time and timestep in ps =
; Apply long range dispersion corrections for Energy and Pressure =
tinit = 0
DispCorr = EnerPres
dt = 0.0001
; Spacing for the PME/PPPM FFT grid =
nsteps = 100000
fourierspacing = 0.12
; number of steps for center of mass motion removal =
; EWALD/PME/PPPM parameters =
nstcomm = 100
pme_order = 4
ewald_rtol = 1e-05
; OUTPUT CONTROL OPTIONS =
; Output frequency for x, v and f=
data logging epsilon_surface = 0
optimize_fft = no
nstxout = 1000 (output files)
nstvout = 0
nstfout = 0
; OPTIONS FOR WEAK COUPLING ALGORITHMS = thermostat
; Temperature coupling =
; Output frequency for energies =
tcoupl = V-rescale
nstlog = 1000
; Groups to couple separately =
nstenergy = 1000
tc-grps = system
; Output frequency and precision for xtc file =
;Time constant (ps) and reference temperature (K) =
nstxtcout = 50000
tau_t = 0.1
xtc-precision = 1000
ref_t = 300
; Pressure coupling =
; NEIGHBORSEARCHING PARAMETERS = barostat
cutoff-scheme = Verlet
cutoff radius Pcoupl = Berendsen
Pcoupltype = semiisotropic
nstlist = 10
; Time constant (ps), compressibility (1/bar) and reference P (bar)
; ns algorithm (simple or grid) =
=
ns_type = grid
tau_p = 1.0
; nblist cut-off =
compressibility = 0 4.5e-5
rlist = 1.0
ref_p = 1.0 1000.0
; Periodic boundary conditions: xyz or none =
; GENERATE VELOCITIES FOR STARTUP RUN =
pbc = xyz
gen_vel = yes
initial velocities
periodic boundaries
gen_temp = 300
...
gen_seed = -1 21
 http://manual.gromacs.org/documentation/5.1/user-guide/mdp-options.html

Simulation settings – Example: NpT simulation

...
o VARIOUS PREPROCESSING OPTIONS =
title = NPT
cpp = /lib/cpp
; OPTIONS FOR ELECTROSTATICS AND VDW = cutoff radius
; Method for doing electrostatics =
include = -I../top
coulombtype = PME
define =
rcoulomb = 1.0
; Method for doing Van der Waals =
; RUN CONTROL PARAMETERS =
integrator = md
MD integrator vdw-type = PME
rvdw = 1.0
; start time and timestep in ps =
; Apply long range dispersion corrections for Energy and Pressure =
tinit = 0
DispCorr = EnerPres
dt = 0.0001
; Spacing for the PME/PPPM FFT grid =
nsteps = 100000
fourierspacing = 0.12
nstcomm
Bottom line
; number of steps for center of mass motion removal =
= 100
; EWALD/PME/PPPM parameters =
pme_order = 4
ewald_rtol = 1e-05
▪ Great number of simulation parameters
; OUTPUT CONTROL OPTIONS =
; Output frequency for x, v and f=
data logging epsilon_surface = 0
optimize_fft = no
nstxout = 1000 (output files)
▪ Need to go deep into theory to understand them all
nstvout = 0
nstfout = 0
; OPTIONS FOR WEAK COUPLING ALGORITHMS = thermostat
▪ Setting up a completely new simulation is a task for a “professional”
; Output frequency for energies =
; Temperature coupling =
tcoupl = V-rescale
nstlog = 1000
▪ For simulations of typical chemical engineering systems, you usually do not
nstenergy = 1000
; Groups to couple separately =
tc-grps = system
; Output frequency and precision for xtc file =
nstxtcout
have to adjust many of them. Allows for using a template like this one.
= 50000
;Time constant (ps) and reference temperature (K) =
tau_t = 0.1
xtc-precision = 1000
ref_t = 300
; Pressure coupling =
; NEIGHBORSEARCHING PARAMETERS = barostat
cutoff-scheme = Verlet
cutoff radius Pcoupl = Berendsen
Pcoupltype = semiisotropic
nstlist = 10
; Time constant (ps), compressibility (1/bar) and reference P (bar)
; ns algorithm (simple or grid) =
=
ns_type = grid
tau_p = 1.0
; nblist cut-off =
compressibility = 0 4.5e-5
rlist = 1.0
ref_p = 1.0 1000.0
; Periodic boundary conditions: xyz or none =
; GENERATE VELOCITIES FOR STARTUP RUN =
pbc = xyz
gen_vel = yes
initial velocities
periodic boundaries
gen_temp = 300
...
gen_seed = -1 22
File structure and information flow in GROMACS
analysis.xtc

Analysis

main_simulation.gro main_simulation.trr main_simulation.edr

Main Simulation

main_simulation.tpr minimization.gro minimization.trr more files ...

Energy minimization
simulation
Step 5 – Energy minimization

minimization.tpr Step 4 – Merge files

P P
sim_settings.mdp
P topology.top minim_settings.mdp
P box_file.gro

oplsaa.ff/forcefield.itp
P MEG.itp
P MEG.pdb
P
ffnonbonded.itp
P P
ffnonbonded.itp reference to file
combination of files or
information flow

Molecular Dynamics Simulations in GROMACS 11.12.2017 Jan Schulze 23

Step 5 – Energy minimization

▪ Molecules randomly inserted by GROMACS, strong overlapping atom shells avoided

▪ Arrangement of molecules can still be problematic
▪ Motivates energy minimization – “Allow molecules to make themselves comfortable.”
▪ Avoids “kicks” and thereby errors at beginning of simulation, i.e. strong repulsive forces

Molecular Dynamics Simulations in GROMACS 11.12.2017 Jan Schulze 24

File structure and information flow in GROMACS
analysis.xtc

Analysis

main_simulation.gro main_simulation.trr main_simulation.edr

Main Simulation

GROMACS generates
main_simulation.tpr minimization.gro minimization.trr more files ...
a lot of output files

Energy minimization
simulation

minimization.tpr

P P
sim_settings.mdp
P topology.top minim_settings.mdp
P box_file.gro

oplsaa.ff/forcefield.itp
P MEG.itp
P MEG.pdb
P
ffnonbonded.itp
P P
ffnonbonded.itp reference to file
combination of files or
information flow

Molecular Dynamics Simulations in GROMACS 11.12.2017 Jan Schulze 25

Important output files

Subsequent
Analysis
simulation

Atom coordinates at Atom trajectories

last instant of time x(t) , v(t)

simulation.gro simulation.trr more files ...

Simulation

simulation.tpr

topology.top settings.mdp box_file.gro

Molecular Dynamics Simulations in GROMACS 11.12.2017 Jan Schulze 26

File structure and information flow in GROMACS
analysis.xtc
▪ Calculate macroscopic
properties from atom
Visualization
Analysis trajectories.
▪ E.g. density and other
thermodyn. properties
main_simulation.gro main_simulation.trr more files ...
▪ Our system is equilibrated!

Main simulation hours, days, weeks

main_simulation.tpr minimization.gro minimization.trr more files ...

Energy minimization seconds, minutes

simulation

minimization.tpr

P P
sim_settings.mdp
P topology.top minim_settings.mdp
P box_file.gro

oplsaa.ff/forcefield.itp
P MEG.itp
P MEG.pdb
P
ffnonbonded.itp
P P
ffnonbonded.itp reference to file
combination of files or
information flow

Molecular Dynamics Simulations in GROMACS 11.12.2017 Jan Schulze 27

Main steps in a GROMACS simulation

1. Define molecular
structure of components

2. Set up force field,

make topology

3. Make box(es) and 5. Perform an energy

insert molecules minimization

4. Merge all necessary 6. Run the simulation

files

7. Check for equilibration,

analyze results

Molecular Dynamics Simulations in GROMACS 11.12.2017 Jan Schulze 28

End

Thanks for your attention!

Questions?

Molecular Dynamics Simulations in GROMACS 11.12.2017 Jan Schulze 29

Step 5 – Energy minimization

White: Before energy minimization

Blue: After energy minimization (you can see periodic boundary conditions for two molecules)

▪ Potential energy can be reduced up to factor 104 depending on system

▪ Avoids “kicks” and thereby errors at beginning of simulation, i.e. strong repulsive forces

Molecular Dynamics Simulations in GROMACS 11.12.2017 Jan Schulze 30

Thermostat: Berendsen coupling

▪ Equations of motion for velocities and positions of the atoms are not only influenced by
interactions but also coupled with control algorithms for pressure and temperature.

▪ Implement a temperature control that realizes a first order behavior:

𝑑𝑇 𝑇0 − 𝑇
=
𝑑𝑡 𝜏

with set-point temperature 𝑇0 and time constant 𝜏 (tuning parameter).

▪ “Strength of coupling”, i.e. aggressiveness of temperature controller can be varied through 𝜏.

 high for fast equilibration

 small during equilibrium simulation subsequent to equilibration process

▪ Time-dependent temperature coupling factor (velocity rescaling factor):

1/2
𝑛 𝑇𝐶 ⋅ Δ𝑡 𝑇0 Δ𝑡 Δ𝑡 𝐅 t
𝜆 𝑡 = 1+ ⋅ −1 ⇒ 𝒗 𝑡+ =𝜆⋅ 𝒗 𝑡− + ⋅ Δ𝑡
𝜏𝑇 𝑇 𝑡 ⋅ 𝑡 − 0.5 ⋅ Δ𝑡 2 2 m

> 1, if T < T0
< 1, if T > T0

Molecular Dynamics Simulations in GROMACS 11.12.2017 Jan Schulze 31