Molecular Modelling

Nguyen Thi Van Oanh

Institut de Chimie Physique
Paris-Saclay University
van-oanh.nguyen-thi@universite-paris-saclay.fr
 Lecture 1
Quantum Mechanics
Quantum Mechanics Simulation
 Quantum Mechanics
̂
Schrodinger Equation: Hψ = Eψ
Ĥ : Hamiltonian operator
ψ: wave function
E: energy eigenvalue

Hamiltonian operator, Ĥ , acts upon the wavefunction, ψ, to generate the evolution of the wavefunction

in time and space

The Schrodinger equation gives the quantized energies, E, of the system and gives the form of the

wavefunction so that other properties may be calculated.

The wavefunction ψ(R, r) is a function of nuclear coordinates R and electron coordinates r

 Method to solve a Schrodinger Equation
Schr dinger Equation can only be solved exactly for simple systems
Examples: Rigid Rotor, Harmonic Oscillator, Particle in a Box, Hydrogen Atom

For more complex systems (i.e. many electron atoms/molecules) we

need to make some simplifying assumptions/approximations and solve it
numerically.
LCAO: linear combination of atomic orbitals
Hartree-Fock approximation
DFT: not necessary to consider the motion of each
individual electron: it suf ces to know the average number
of electrons located at any one point in space
Used now in Gaussian-90 to calculate the geometrical
structure of molecules and map chemical reactions.
Drawback: time-consuming for very large biomolecules
ö
fi
 Simulation Packages

There exist a large number of software packages capable of performing

electronic structure calculations.
For examples: MOLPRO, GAMESS, COLUMBUS, NWCHEM, MOLFDIR, ACESII,
GAUSSIAN, ...

The different programs have various advantages and capabilities.

 Applications

Calculate the potential surface of the isomerisation

Photo-switches are one type of molecular machines, a class of molecules that can be
switched between at least two distinct thermodynamically stable forms by the
application of an external stimulus. hence are ideal tools for different research areas
spanning from chemical biology to smart materials
Applications
 Applications
Biochemistry
 Lecture 2
Molecular Mechanics
Molecular mechanics methods are based on the following principles:
・Nuclei and electrons are lumped into atom-like particles

Atom-like

・Atom-like particles are spherical (radii obtained from measurements or theory) and

have a net charge (obtained from theory).

・Interactions are based on classical potentials.

・Interactions must be preassigned to speci c sets of atoms.

・Interactions determine the spatial distribution of atom-like particles and their energies.
fi
 Bond stretch energy
• The bond stretch term describes energy changes of bond lengths with respect to their
equilibrium values

• Force eld parameters: ai and lio

Examples:

• C-C bond:
ai= 50 kcal/mol/Å2
lo =1.5 Å2

• C=C bond:
ai= 100 kcal/mol/Å2
lo =1.3 Å2

• The force eld parameters of bond stretch are usually parametrised from spectroscopy
fi
fi
 Bond angle energy

• The bond angle term describes energy changes of bond angles with respect to their
equilibrium values

• Force eld parameters: bi and θio

• The force eld parameters of bond angle are usually parametrised from spectroscopy
fi
fi
 Torsional angle energy

• The torsional angle term describes energy changes associated with rotations around
chemical bond

• Force eld parameters: ci and γi and n

• The force eld parameters of torsional angle are usually parametrised from spectroscopy
fi
fi
 Van der Waals energy

• The van der Waals term describes the repulsive and attractive inter-atomic forces.
• Force eld parameters: εij and σij

• The force eld parameters of van der Waals are usually parametrised from
small molecular crystals
fi
fi
 Electrostatic energy

• The electrostatic term describes Coulomb electrostatic potential between

charged atoms

• Force eld parameters: qi and qj

• The force eld parameters of electrostatic are usually parametrised from

ab initio quantum mechanics ( t potential and/or eld)
fi
fi
fi
fi
 Classical Potential Functions
• The total potential energy of a system is the sum of all energy terms

U= +
+

+ +

+
 The Force Field
These total potential functions together with the data (parameters) required to describe
the behavior of different kinds of atoms and bonds, is called a “FORCE FIELD”.

The molecular mechanics “FORCE FIELD” relates the motions, and energies of
motions of atoms within the molecule. The force eld is used to govern how the
parts of a molecule relate to each other, that is, how each atom or group of atoms is
affected by its environment, and how these forces contribute to the structure of the
molecule.

Many different kinds of force- elds have been developed over the years.
Some include additional energy terms that describe other kinds of deformations.
Some force- elds account for coupling between bending and stretching in adjacent
bonds in order to improve the accuracy of the mechanical model.
fi
fi
fi
 Popular molecular mechanics force elds
AMBER (Assisted Model Building and Energy Re nement) - widely used for proteins
and DNA

CHARMM - originally developed at Harvard, widely used for both small molecules and
macromolecules

CHARMm - commercial version of CHARMM, available through Accelrys CVFF - also

broadly used for small molecules and macromolecules

GROMACS - The force eld optimized for the package of the same name GROMOS
- A force eld that comes as part of the GROMOS (GROningen MOlecular Simulation
package), a general-purpose molecular dynamics computer simulation package for the
study of biomolecular systems.

GROMOS force eld (A-version) has been developed for application to aqueous or
apolar solutions of proteins, nucleotides and sugars. However, a gas phase version (B-
version) for simulation of isolated molecules is also available

OPLS-aa, OPLS-ua, OPLS-2001, OPLS-2005 - Members of the OPLS family of

force elds developed by William L. Jorgensen at Yale Department of Chemistry.
fi
fi
fi
fi
fi
fi
Popular molecular mechanics force elds

fi
 Force Fields Parameters
In addition to the functional form of the potentials, a force eld typically de nes a set
of parameters for each of a number of atom or particle types that correspond to
different atoms and bonding patterns in commonly simulated molecules.

The parameter set includes values for atomic mass and partial charge for individual
atoms, and equilibrium bond lengths and angles for pairs, triplets, and quadruplets of
bonded atoms.

Example:
AMBER force eld
parameters of Phenylalanine
residue
fi
fi
fi
 Molecular Mechanics Force Fields:
Parameterization

Model compounds: Although many molecular simulations involve biological

macromolecules such as proteins, DNA, and RNA, the parameters for given atom
types are generally derived from observations on small organic molecules that
are more tractable for experimental study and quantum calculation.

Quantum calculation data: optimized geometries, conformational energies, interaction

energies, electric moments, electrostatic potentials, electron densities

Experimental data: crystal/NMR structures, vibrational frequencies, pure liquid/solid

properties, solvation free energies, NMR data such as J-J couplings, order parameters
 Strategies of AMBER Force Field Parameterization

1. PartialCharge
HF/6-31G* RESP (derive charges to reproduce ab initio ESP)

2. Van der Waals

Reproduce bulk properties, such as density and heat of vaporization, hydration free
energies

3. Bond length and bond angle parameters

Experiments or high-level ab initio calculations

4. Torsional angle parameters

Experimental or high-level ab initio relative energies and rotational pro les

fi
Force Field Comparison: Examples
Force Field Comparison: Examples
Force Field Comparison: Examples
 Lecture 3
Polarizable Molecular Mechanical Force Field
Polarizable Molecular Mechanical Force Field
Polarizable Molecular Mechanical Force Field
Polarizable Molecular Mechanical Force Field
The use of Coulomb’s law with xed atomic charges to treat the electrostatic
interactions is a major simpli cation in current force elds.
Describe electrostatic interaction in a more physical fashion
Suitable to study heterogeneous systems, dielectric continually changed
system, highly charged systems.

0
∑
Vpol = − 1/2 μpEp
p
fi
fi
fi
Some Polarizable Force Fields
Comparison of Polarizable and Additive Force Fields
Comparison of Polarizable and Additive Force Fields
 Lecture 5
Implicit Solvent
 Why Solvent ?
Many reactions take place in solution
Consider

Water models:
SPC/E, TIP3P, TIP4P, TIP5P
Can we ignore solvent in simulations?
We can’t ignore solvent, but consider solvent implicitly
Strategy to construct an implicit solvent model

It is possible to construct an implicit solvent model by approximating

the medium outside the water-excluded volume as a continuum with
electrostatic, entropic, and viscous properties that match water.

Molecular Mechanics Atomistic Force Fields With Solvent

Effect Taken Into Account
Strategy to construct an implicit solvent model
It is possible to construct an implicit solvent model by approximating
the medium outside the water-excluded volume as a continuum with
electrostatic, entropic, and viscous properties that match water.
Strategy to construct an implicit solvent model

It is possible to construct an implicit solvent model by approximating

the medium outside the water-excluded volume as a continuum with
electrostatic, entropic, and viscous properties that match water.

Molecular Mechanics Atomistic Force Fields With Solvent

Effect Taken Into Account
Strategy to construct an implicit solvent model
 Model the Non-Polar Component: ΔGcav + ΔGdisp

Solvent-Accessible Surface Area method

The free energy of the solvation free energy component

ΔGnon−polar of a solute molecule is described by a
SASA
mean solvation potential Vsolv .

It is computed as the product of an atom-speci c solvation

energy per surface area σiSASA and the atomic SASAi, summed over all atoms i.
SASA
σiSASA . SASAi( r ⃗)
∑
Vsolv ( r ⃗) =
i
fi
Model the Polarisation Component: ΔGelec
 Model the Polarisation Component: ΔGelec
Poisson-Boltzmann Method (accurate but numerical and slow)
 Generalized Born Approximation
Generalized Born models (faster, can be expressed as analytic function)
 Generalized Born (GB) Approximation
In the GB method a pseudo-ideal situation is modelled locally in non-ideal solutes

(like biomolecules) by variation of ri

By xing the (computed or measured) value of ΔGelec, the undetermined radius r

can be varied as control parameter to match the given ΔGelec. The resulting
‘effective Born radius’ ri adjusts the local screening to an optimal value modulo
approximations of the GB theory. The effective Born radii are used for the
computation of the GB pair terms between charged atoms or atom groups i and j:

1 1 1 qiqj
ΔGelec = − ( −
2 ϵin ϵout ) rij2 + rirj exp(−rij2 /4rirj)

Most major biomolecular simulation packages (CHARMM, Amber, IMPACT,

Gromacs, etc.) include pairwise descreening GB implementations suitable for MD
calculations.
fi
 Lecture 6
Coarse-Grained Molecular Force Field
 Coarse-grained models
• In the all-atom (AA) model, all atoms are considered explicitly
• In the coarse-grained (CG) model, small groups of atoms are
represented by coarse-grained beads.
The Coarse-grained MARTINI Model
 The Protein Lattice Model
We used the Frenkel lattice model Plos ONE 2014

Here we used Estate and Esolvent=0, and two-body and four-body Ehb (OPEP)
The Coarse-grained OPEP Model
 The Coarse-grained OPEP Model
PEP-FOLD 1 RESULTS

52 peptides and 13 miniproteins of 9-23 and 27-49 amino acids: RMSd = 2.3 and 3.7 Å
Maupetit et al., Nucleic Acid Res (2009) and J. Comput Chem. (2010)
 Lecture 7
Molecular Dynamics Simulation
Molecular dynamics simulation

• No explicit electrons
• No polarisation, electron transfer or correlation
• Conformational energies for ground state
• No chemistry
• Not variational bound
Molecular dynamics simulation: the Principle

Newton equation: F = m.d2r/dt2

Force Mass Acceleration

Solve Newton equation numerically by computer to predict

next position from a previous position
 Molecular dynamics simulation
• Let’s consider a molecule containing N atoms with coordinates
r={ri}, where ri=(rix,riy,riz) denotes the position of the i-th atom.
• Total potential: U = Vbond + Vangle + Vdihedral + VCoulomb + VvanderWaals

Newton’s equation of motion

 Molecular dynamics simulation: the Force Field

• We have to de ne the interaction between atoms of the system

• An empirical set of potential functions and associated constants to describe such

interaction is called force eld

Potential functions

Bonded interactions
Non-bonded interactions
Bond stretch

van der Waals

Bond angle

Electrostatic
Torsional angle
fi
fi
 Bond stretch energy
• The bond stretch term describes energy changes of bond lengths with respect to their
equilibrium values

• Force eld parameters: ai and lio

Examples:

• C-C bond:
ai= 50 kcal/mol/Å2
lo =1.5 Å2

• C=C bond:
ai= 100 kcal/mol/Å2
lo =1.3 Å2

• The force eld parameters of bond stretch are usually parametrised from spectroscopy
fi
fi
 Bond angle energy

• The bond angle term describes energy changes of bond angles with respect to their
equilibrium values

• Force eld parameters: bi and θio

• The force eld parameters of bond angle are usually parametrised from spectroscopy
fi
fi
 Torsional angle energy

• The torsional angle term describes energy changes associated with rotations around
chemical bond

• Force eld parameters: ci and γi and n

• The force eld parameters of torsional angle are usually parametrised from spectroscopy
fi
fi
 Van der Waals energy

• The van der Waals term describes the repulsive and attractive inter-atomic forces.
• Force eld parameters: εij and σij

• The force eld parameters of van der Waals are usually parametrised from
small molecular crystals
fi
fi
 Electrostatic energy

• The electrostatic term describes Coulomb electrostatic potential between

charged atoms

• Force eld parameters: qi and qj

• The force eld parameters of electrostatic are usually parametrised from

ab initial quantum mechanics ( t potential and/or eld)
fi
fi
fi
fi
 Total potential energy of the system

• The total potential energy of a system is the sum of all energy terms

U= + +

+ +

• Popular force elds such as OPLS, CHARMM, AMBER have the same
potential functional form, but different in the parameters

• In the TP course, you will use and learn more about the AMBER force eld
fi
fi
 Molecular dynamics
• Having described the potential energy of the system, the force acting on
the i-th atom is calculated from the potential energy as:

• Use this force to calculate the next position of the i-th atom
 The Verlet Algorithm

• Remind: the Taylor expansion: Derivation

• There are also other algorithms

 Start the simulation

• We have idea about molecular dynamics simulation (concept, history)

• We de ne the potential energy for the system (force eld)
• We know how to predict position of atoms at anytime t (Verlet algorithm)
Now, let’s start
The initial coordinates of atoms: x(t=0)
•Taken from Protein Data Bank (PDB)
https://www.rcsb.org
•From previous simulation
•By construction
ATOM 1 N ALA 1 13.670 11.700 11.510 1.00 0.00
ATOM 2 H1 ALA 1 13.020 10.950 11.400 1.00 0.00
ATOM 3 H2 ALA 1 14.340 11.460 12.210 1.00 0.00
ATOM 4 H3 ALA 1 13.170 12.520 11.790 1.00 0.00
ATOM 5 CA ALA 1 14.370 11.950 10.220 1.00 0.00
ATOM 6 HA ALA 1 14.980 11.070 10.020 1.00 0.00
ATOM 7 CB ALA 1 13.370 12.190 9.060 1.00 0.00
ATOM 8 HB1 ALA 1 12.590 11.440 8.940 1.00 0.00
ATOM 9 HB2 ALA 1 13.890 12.230 8.100 1.00 0.00
……
fi
fi
 Energy minimisation

• Our molecule may have steric clashes, inappropriate geometry, atoms are not
proper arranged…, therefore molecular structure is not energetically favourable.

• Before beginning dynamics, energy minimization ensures that the system nds
appropriate starting structure. The molecule is relaxed during this process

Our initial structure is here

…but we want to start

simulation from here

fi
Energy minimisation
 Energy minimisation

Demonstration: nd local minimum using rst order derivative

fi
fi
Energy minimisation: steepest descent
Energy minimisation: conjugate gradient
 Mimic experimental conditions

• Suppose that an experiment was done at temperature T=300 K. How can the simulation
mimic this experimental condition?

==> The initial velocity of atoms is generated from the Maxwellian distribution for the
required temperature T=300 K
 Mimic experimental conditions: thermostat

•Temperature during the simulation T(t) is maintained by Thermostat (temperature

control)

Thermostat:
• The velocity of each atom is rescaled during the simulation (Berendsen thermostat):

Δt T(t)
Vnew = vold (1 + [ − 1])
τ T0

•Popular thermostats:

Berendsen thermostat,
Nose-Hoover thermostat
Bussi-Donadio-Parrinello thermostat

You will play with parameters: T0, τ

and different thermostat in the TP
 Mimic experimental conditions: barostat

•Initial pressure P0 is selected according to the experimental pressure

•Pressure during the simulation P(t) is maintained by Barostat (pressure control)

Barostat:
• The simulation box sizes are rescaled during the simulation (Berendsen thermostat)

βΔt
Lnew = Lold(1 − [P(t) − P0])1/3
τp

•Popular thermostats:

Berendsen thermostat
Parrinello-Rahman thermostat

You will play with parameters: P0, τp

and different barostats in TP
 Mimic experimental conditions: PBC

PBC: Periodic Boundary Condition

• We are interested in experimental results done in bulk environment which are not
affected by the system surfaces.
• Surface effects are eliminated using the so called "periodic boundary conditions”
• PBC makes it possible to approximate an in nite system by using a small part (unit cell)
No PBC With PBC

• A particle that leaves the cell on the one side is replaced by a copy entering the cell
on the opposite side.
fi
Treat long-range interactions
Treat long-range interactions
Choice of time step
 Method for increasing time step

• One way to increase the time step is to eliminate the fastest motions from the system.
• The so-called SHAKE algorithm is widely used to effectively x fast vibrating bonds
and angles at their equilibrium values.

fi
 Lecture 8
Enhanced Sampling Methods
Conformational sampling: the problem

The molecule is trapped

• we misss some structures
• incorrect thermodynamics ensemble
trapped
here
At 300 K:
KBT ≈ 2.5 kJ/mol

5 - 20 kJ/mol
Good conformational sampling

The molecule should sample all possible states

 How to obtain good sampling?
• The molecule has to pass through all possible conformational states
to calculate accurately observables.
Strategy 1: Let the molecule to evolve in long time (dif cult)

❖ Special Anton supercomputer: folding

simulations on μs - ms timescales
(Science 334, 517, 2011)

Anton is a massively parallel supercomputer designed and built

by D. E. Shaw Research. It is a special-purpose system for
molecular dynamics simulations of proteins and other biological
macromolecules

fi
 How to obtain good sampling?
Strategy 2: Coarse-grained models

• Quantum mechanical (QM) model: treats every electron (very costly)

• All-atom (AA) model: treats every atom (costly)
• Coarse-grained (CG) model: combines several atoms into one single bead
(hundreds of times faster)
• Dissipative particle dynamics (DPD) model:
coarse-grained degree is much larger
(millions of times fasters)
• Continuum models: represents processes
by differential equations

Combine several models:

multi-scale simulation
Strategies to model complex systems
• Construction of a model of the system and (more or less) exact
treatment of dynamics: Quantum-mechanical models

• Detailed microscopic description of system and approximate

treatment of dynamics: Classical-mechanical models

ails
det

ultra coarse-grained

coarse-grained
cale
m e s
, t i
size
tem
all-atom sys
 How to obtain good sampling?
Strategy 3: Accelerate the motion of molecule (how?)

Enhanced Sampling Methods

We need something from outside to help sampling
Temperature Replica-Exchange: implementation
❖ Generate N copies (replicas) of the system
❖ Exchange replica pairs every n MD steps

500 K Rep3
temperature

exchange MD
400 K Rep2
MD
exchange
300 K Rep1
MD
simulation time

• Ideal for parallel computing, easy to program

• Not suitable for large systems (N ∼ √f)
 Temperature Replica-Exchange: an example
random walk in temperature

normal sampling
enhanced sampling

P. H.Nguyen, Y. Mu and G. Stock. Proteins 60:485 (2005)

Simulated tempering method: the algorithm
1. Run MD for i steps at temperature βm and weight fm
2.Update the state (βm,fm) --> (βn,fn) with probability:

w(βm -> βn) = min(1,e-[(βn-βm)Em - (fn-fm)])

500 K f3
temperature

e c t
re j
exc
e 400 K f2 400 K f2 ha
ang MD MD
ng
e
ch
ex
300 K f1 400 K f2
MD MD

simulation time
At one temperature, ST is just a normal MD: no problem with large
systems...
but how to choose fm prior the simulation ?
 Simulated tempering: 37-residue WW domain

Simulated tempering

• AMBER99SB*-ILDN
Starting • 48 temperatures: 340-485 K
• ST: 1 μs

Using Anton machine

Lindorff-Larsen K, Piana S, Dror RO, Shaw DE
Science 334, 517 (2011)

Tong et al. J. Phys. Chem. B 119, 6941 (2015)

Simulated tempering:20-residue Tap-Cage peptide

Simulated tempering
• AMBER99SB*-ILDN
Starting • 20 temperatures: 310-400 K
• ST: 1 μs

Using Anton machine

Lindorff-Larsen K, Piana S, Dror RO, Shaw DE
Science 334, 517 (2011)

Tong et al. J. Phys. Chem. B 119, 6941 (2015)

Metadynamics: the idea

+ =
 Metadynamics
•Metadynamics is a technique for enhancing conformational sampling in
molecular dynamics simulations along some collective variables.
A. Laio & M. Parrinello. Proc. Natl. Acad. Sci. 99, 12562 (2002)

• What is a collective variable ?

- Any function of any number of variables: position, distance,
angle, dihedral, density, NMR spectrum etc.
- Whatever you would like to explore, but it should be able to
describe the activated process of interest.

For example: dihedral angles of proteins

 Metadynamics: the idea
System: to demonstrate the idea, consider the
dynamics of the system on a 2D potential energy
surface U(x,y). The transition between minima is rare
Collective variable: to accelerate this transition,
we can bias the system along the x-axis, and
our collective variable is: s(x) = x
Biased potential: we construct an external
history-dependent bias potential as a function of s:

Recall: The normal dynamics is biased by the

metadynamics potential which pushes
the system away from local minima into
visiting new regions of the phase space
 Metadynamics: illustration

The normal dynamics is biased by the metadynamics

potential which pushes the system away from local
minima into visiting new regions of the phase space
 Metadynamics: an example
Usually, one consider several collective variables si to describe the
dynamics of the system. The biased potential is generalized as:

8 collective variables are used:

coordination number, antiparallel rmsd, parallel rmsd, number of antiparallel β-
sheets, number of parallel β-sheets, number of antiparallel steric zippers.
 Lecture 9
Data Analysis
 End
08 March 2024
Microstate

Each microstate occurs with a given probability Pi

 Ensemble of microstates

• An ensemble is a collection of microstates that are compatible with a speci ed

macrostate of a thermodynamic system

fi
The partition function

Boltzmann factor
 An example

Let’s consider a 2-state system.

Potential

The probability to be each state:

U1 exp(−Ui /kBT )
Pi =
U0 2
∑i exp(−Ui /kBT )
2 kcal/mol
0
r
exp(−0/kBT ) exp(−2/kBT )
P1 = P2 =
exp(−0/kBT ) + exp(−2/kBT ) exp(−0/kBT ) + exp(−2/kBT )

At temperature T = 300 K, kBT = 0.6 kcal/mol. What are the values of P1 and P2?
Macrostate
Average of a quantity in practice
 Brief Theory of Monte Carlo Simulation

Idea: generate a series of con gurations by comparing energies, no forces calculated

Potential

Potential
Always accept accept with probability

E E’
E’ E
0 0
r r
Transition probability from state E to state E’:

1 if E′ − E ⩽ 0,
{e
W(E − > E′) = −(E′ − E)
kBT
if E′ − E > 0




fi
 Metropolis criterion

The Metropolis criterion is used to decide if a new con guration is accepted or not

fi
Molecular dynamics versus Monte Carlo
Practice Course
 Molecular dynamics simulation with GROMACS

Currently, there are several free or low‐cost software packages for

performing MD simulations

Software package Cost Website

Desmond Free www.deshawresearch.com/resources_desmond.html

Gromacs Free www.gromacs.org
NAMD Free www.ks.uiuc.edu/Research/namd
Abalone Free www.biomolecular-modeling.com/Abalone
AMBER $500 ambermd.org
CHARMM $600 www.charmm.org
 Molecular dynamics simulation with GROMACS

Why GROMACS?
(Groningen Machine for Chemical Simulations)

GROMACS own manual says about itself:

GROMACS is a versatile package to perform molecular dynamics, i.e. simulate

the Newtonian equations of motion for systems with hundreds to millions of
particles.
It is primarily designed for biochemical molecules like proteins, lipids
and nucleic acids that have a lot of complicated bonded interactions, but
since GROMACS is extremely fast at calculating the nonbonded interactions
(that usually dominate simulations) many groups are also using it for
research on non-biological systems, e.g. polymers.
 Case Study: Trialanine

• Trialanine consists of three alanine amino-acids.

• Trialanine represents an ideal model system because it is among the simplest

systems that exhibit typical features of biomolecules.

• Trialanine is small enough allowing you to understand the structure of the

peptide in details, to finish the simulation and to analyze common but
important quantities within a reasonable time.

• All basic techniques which you learn from this simulation can be used to
simulate any other systems.
Basic steps
of MD
 MD simulation of Trialanine

Step 1: Directory organisation

Create a directory to store all files of the first trialanine system:

mkdir ALA3

Step 2: Obtain starting structure

The starting structure file (ala3.pdb) has been provided as part of the
tutorial package. Please copy this file to the ALA3 directory:

cp ala3.pdb ./ALA3

Atom number Element Residue Atom coordinates (x,y,z)

ATOM 1 N ALA 1 13.670 11.700 11.510 1.00 0.00

ATOM 2 H1 ALA 1 13.020 10.950 11.400 1.00 0.00
ATOM 3 H2 ALA 1 14.340 11.460 12.210 1.00 0.00
ATOM 4 H3 ALA 1 13.170 12.520 11.790 1.00 0.00
ATOM 5 CA ALA 1 14.370 11.950 10.220 1.00 0.00
ATOM 6 HA ALA 1 14.980 11.070 10.020 1.00 0.00
ATOM 7 CB ALA 1 13.370 12.190 9.060 1.00 0.00
ATOM 8 HB1 ALA 1 12.590 11.440 8.940 1.00 0.00
ATOM 9 HB2 ALA 1 13.890 12.230 8.100 1.00 0.00
……
 MD simulation of Trialanine
Step 3: Select Force Field and generate topology file

•Set-up of the force field

•Generation of the “topology”
•Tool: gmx pdb2gmx of GROMACS:
gmx pdb2gmx -f ala3.pdb -o ala3.gro -p ala3.top -i ala3.itp -ignh

Input Output Output

coordinate file coordinate file topology file
; residue 1 ALA rtp NALA q +1.0
1 N3 1 ALA N 1 0.1414 14.01 ; qtot 0.1414
2 H 1 ALA H1 2 0.1997 1.008 ; qtot 0.3411
……….

[ bonds ]
Structure of the topology le
; ai aj funct c0 c1 c2 c3 Indices for calculation of bond
1 2 1
stretching potential energy
1 3 1

[ pairs ]
; ai aj funct c0 c1 c2 c3
1 8 1 1-4 pairs for Lennard-Jones and
1 9 1 Coulomb interactions

[ angles ]
; ai aj ak funct c0 c1 c2 c3
2 1 3 1 Indices for calculation of angle
2 1 4 1 bending potential energy
[ dihedrals ]
; ai aj ak al funct c0 c1 c2 c3 c4
2 1 5 6 9
2 1 5 7 9 Indices for calculation of dihedral
[ dihedrals ]
angle bending potential energy
; ai aj ak al funct c0 c1 c2 c3
5 13 11 12 4
11 15 13 14 4 Indices for calculation of improper
15 23 21 22 4
dihedral angle bending potential
; Include Position restraint file energy
#ifdef POSRES
#include "posre.itp"
#endif

; Include water topology

#include "amber99sb-ildn.ff/tip3p.itp"
Topology file for water molecule

#ifdef POSRES_WATER
; Position restraint for each water oxygen
[ position_restraints ]
; i funct fcx fcy fcz
1 1 1000 1000 1000
#endif

; Include topology for ions

#include "amber99sb-ildn.ff/ions.itp"

[ system ]
; Name
Generated by trjconv : Protein in water t= 0.00000
[ molecules ]
; Compound #mols
Protein 1
fi
 MD simulation of Trialanine

Step 4: Create a simulation box

•Create a box around our peptide that we can later fill with solvent.
•The peptide is centered in the simulation box with a minimum distance to
any wall of 1.0 nm.

•Tool: gmx editconf of GROMACS

gmx editconf -f ala3.gro -o ala3-box.gro -bt cubic -d 1.0

Input Output
Coordinate file Coordinate file

You can have a look at the box with VMD by loading:

vmd ala3-box.gro

and type “pbc box” in the VMD console.

The box contains a trialanine molecule.

 MD simulation of Trialanine
Step 5: Adding water to the box

•Add water to the box that we just generated using the command
•Tool: gmx solvate

gmx solvate -cp ala3-box.gro -cs spc216.gro -o ala3-solvated.gro -p ala3.top

Input coordinate Output coordinate Output topology

file without water file with water file with water

You can have a look at the box with VMD by loading:

vmd ala3-solvated.gro

Open the topology file (ala3.top) you will see the

number of water molecules at the last line

The box contains a trialanine

molecule (green) surrounded by
waters.
 MD simulation of Trialanine

Step 6: Define simulation parameters

•Define all parameters:

- Energy minimiser
- Thermostat, barostat,
- Integration algorithm,
- Number of MD steps
- Temperature, pressure.

•All parameters are listed in a .mdp file (files em.mdp and eq.mdp provided).
 MD simulation of Trialanine
Step 7: Energy minimization

•This will relax some unnatural distances between atoms which are
accidentally too close to each other during the setup.

•GROMACS uses the gmx grompp tool to combine the coordinate (ala3-solvated.gro)
and topology (ala3.top) files with parameters from the (em.mdp) file to
generate a single input file (em.tpr) file for the minimization.
 MD simulation of Trialanine
•All files are merged into a single file with tool: gmx grompp

gmx grompp -f em.mdp -c ala3-solvated.gro -p ala3.top -o em.tpr -r ala3-solvated.gro -maxwarn -1

Input Input Input Output

Parameter file coordinate file Topology file file

•The energy minimization is done using tool: gmx mdrun

gmx mdrun -s em.tpr -e em.edr -g em.log -o em.trr -c em.gro

Input Output
file Coordinates file
Output
(Last structure)
Energy file

Output
Log file
Output
Trajectory file
 MD simulation of Trialanine

Exercise 1:
To see how the potential energy of the system changes during the
minimization, you can extract the potential energy from file (em.edr)
using the command energy:

gmx energy -f em.edr -o potential.xvg

and plot the graph with the program xmgrace:

xmgrace potential.xvg

• How does the potential energy behave? Report the initial and final energy values.
•Try with the conjugate gradient method by specifying: integrator = cg in the em.mdp file.
Then, run the energy minimization again, and compare the potential energy with that obtained
by “steep” minimiser. Which algorithm is more efficient?
 MD simulation of Trialanine
Step 8: Equilibration
•Give the system a desired temperature and pressure.
•Relax the simulation box to a desired density
•GROMACS uses the gmx grompp tool to combine the energy minimised coordinate
(em.gro) and topology (ala3.top) files with parameters from the (eq.mdp) file
to generate a single input file for the equilibration.
 MD simulation of Trialanine
Step 8: Equilibration
•Give the system a desired temperature and pressure.
•Relax the simulation box to a desired density
•GROMACS uses the gmx grompp tool to combine the energy minimised coordinate
(em.gro) and topology (ala3.top) files with parameters from the (eq.mdp) file
to generate a single input file for the equilibration.
 MD simulation of Trialanine
•All files are merged into a single file with tool: gmx grompp

gmx grompp -f eq.mdp -c em.gro -p ala3.top -o eq.tpr -r em.gro -maxwarn -1

Input Input Input Output

Parameter coordinate Topology file
file file file

•The equilibration is done using tool: gmx mdrun

gmx mdrun -s eq.tpr -e eq.edr -g eq.log -o eq.trr -c eq.gro

Input
file Output Output
Energy file Coordinates file
Output (Last structure)
Log file
Output
Trajectory file
 MD simulation of Trialanine

•Visualize the temperature, pressure and energy using the gmx energy tool:

gmx energy -f eq.edr -o energy.xvg

gmx energy -f eq.edr -o temperature.xvg
gmx energy -f eq.edr -o pressure.xvg

•Please create plots using xmgrace. Comments on the results.

Exercise 2:

Please repeat the equilibration process, but using different parameters:

1. Run an equilibration for 250 ps.
2. Using the Nose-Hoover thermostat, and Parrinello-Rahman barostat
3. All other parameters are unchanged.
4. Compare the energy, temperature and pressure with those obtained above.
 MD simulation of Trialanine
Step 9: Production

•Run the simulation longer to obtain equilibrium structures of trialanine

•Modify the (eq.mdp) file and modify it, taking into account the needs for
performing a production MD simulation.

Exercise 3: You will run a production simulation with the following requirements:
1. Starting from the equilibrium structure obtained from Step 8.
2. The time length of the trajectory is 1 ns
3. Use the velocity-verlet algorithm for integration
4. Because the simulation box was adjusted to the correct size in the
equilibrium step, we do not need a barostat (turn off barostat).
5. We use the v-rescale thermostat.
6. We need a simulation at room temperature of 300 K.
7. Because atoms already had velocities from equilibration step, we do not
need to generate velocities.

•After modification of file (eq.mdp) taking into account 1-7, you save the
file with name (md.mdp), and follow simulation steps:

gmx grompp -f md.mdp -c eq.gro -t eq.trr -p ala3.top -o md.tpr -r eq.gro -maxwarn -1

and run the MD production simulation:

gmx mdrun -s md.tpr -e md.edr -g md.log -o md.trr -c md.gro

 MD simulation of Trialanine
Analysis of results
Check the equilibration and trajectory visualization

1. Visualize the time evolution of temperature, pressure and total energy:

You use the gmx energy tool to extract these quantities from the (md.edr)
file.

2. Visualize the trajectory:

The trajectory file (md.trr) is written in the binary format. Therefore,
it is convenient to convert the trajectory into a readable format:

gmx trjconv -f md.trr -s md.tpr -o md-traj.gro

You can visualize the trajectory by using the vmd command of VMD:

vmd md-traj.gro

Do you see a strange behavior? Why? Now you try with:

gmx trjconv -f md.trr -s md.tpr -o md-traj.gro -pbc nojump

And visualise it again.

 MD simulation of Trialanine
Analysis of results
Calculation of some structural quantities
1. Calculate the root-mean-squared-deviation (RMSD)
You use the GROMACS gmx rms tool to calculate the RMSD of structures
along trajectory stored in the file md.trr with respect to the initial
structure at t = 0 ps:

gmx rms -f md.trr -s md.tpr -o rmsd

Then visualise file:

xmgrace rmsd.xvg

You will see that the RMSD changes in time, indicating the peptide
explores different structures.

2. Calculate the end-to-end distance

You have to tell GROMACS the indices of these atoms by creating a file
named index.ndx, containing the following lines:

[ distance ]
x y

where x and y are the indices of the first and last Cα atoms in the coordinate file Then, use the gmx dista
tool:

gmx distance -f md.trr -s md.tpr -n index.ndx -oall distance

The generated data is saved in table-form into the distance.xvg file and can be plotted:

xmgrace distance.xvg
 Case Study: Amyloid β peptide

•Initial structure of the peptide is taken from Protein Data Bank (PDB)
https://www.rcsb.org
with the PDB code: 1zOq

• Perform all necessary steps

II. Enhanced Sampling: Replica-Exchange MD method
 Why do we need enhanced sampling methods
bond angle

dihedral angle Complex landscape

Conformational sampling: the problem
The molecule is trapped ==> incorrect thermodynamics ensemble

trapped
here
At 300 K:
KBT ≈ 2.5 kJ/mol

5 - 20 kJ/mol
 Good conformational sampling

At 500 K:
KBT ≈ 4.2 kJ/mol

4 - 10 kJ/mol

The molecule should sample all possible states

 How to obtain good sampling?

We need something from outside to cross the barriers

 Temperature Replica-Exchange: the idea
❖ Generate N copies (replicas) of the system
❖ Exchange replica pairs every n MD steps

500 K Rep3
temperature

exchange MD
400 K Rep2
MD
exchange
300 K Rep1
MD
simulation time

• Ideal for parallel computing, easy to program

• Not suitable for large systems (N ∼ √f)
 Replica exchange simulation with GROMACS

•The first step is to obtain a list of temperatures distributed between

minimum (Tmin) and maximum (Tmax) temperatures

T0 = 300 K
T1 = 302 K

T2 = 305 K
T3 = 309 K
T4 = 313 K

T5 = 318 K
T6 = 323 K
…
…
T20 = 400 K

•Note: the list of temperatures is system-dependent

•How to choose the temperatures: see the lecture
 Replica exchange simulation with GROMACS

•To run REMD with GROMACS, you first create different parameter files which
are only different in temperatures.
•Take the md.mdp file you used in previous simulation, and rename it to md0.mdp,
md1.mdp, md2.mdp and md3.mdp etc. Then change the temperature:

ref-t = 300 300 in file md0.mdp

ref-t = 302 302 in file md1.mdp
ref-t = 305 305 in file md2.mdp
ref-t = 309 309 in file md3.mdp
…..

•For each temperature, you generate the corresponding input file:

gmx grompp -f md0.mdp -c eq.gro -t eq.trr -p ala3.top -o md0.tpr
gmx grompp -f md1.mdp -c eq.gro -t eq.trr -p ala3.top -o md1.tpr
gmx grompp -f md2.mdp -c eq.gro -t eq.trr -p ala3.top -o md2.tpr
gmx grompp -f md3.mdp -c eq.gro -t eq.trr -p ala3.top -o md3.tpr
…
•Then GROMACS runs the REMD simulation using the command:

mpirun mdrun -np 4 -s md.tpr -e md.edr -g md.log -o md.trr -c md.gro -multi 4

 Analyze REMD trajectory
Calculate several quantities of REMD trajectories and compare the results
1. Calculate the root-mean-squared-deviation (RMSD)

gmx rms -f md0.trr -s md0.tpr -o rmsd0

gmx rms -f md1.trr -s md1.tpr -o rmsd1
gmx rms -f md2.trr -s md2.tpr -o rmsd2
gmx rms -f md3.trr -s md3.tpr -o rmsd3

Then visualise file:

xmgrace rmsd0.xvg rmsd1.xvg rmsd2.xvg rmsd3.xvg

You will see that the RMSD changes more at higher temperature replicas,

2. Calculate the end-to-end distance

You have to tell GROMACS the indices of these atoms by creating a file named index.ndx, containing
the following lines:

[ distance ]
x y

where x and y are the indices of the first and last Cα atoms in the coordinate file.
Then, use the gmx distance tool:

gmx distance -f md0.trr -s md0.tpr -n index.ndx -oall distance0

gmx distance -f md1.trr -s md1.tpr -n index.ndx -oall distance1
gmx distance -f md2.trr -s md2.tpr -n index.ndx -oall distance2
gmx distance -f md3.trr -s md3.tpr -n index.ndx -oall distance3

The generated data is saved in table-form into the distance.xvg file and can be plotted:

xmgrace distance0.xvg distance1.xvg distance2.xvg distance3.xvg

What do you observe?

 III. Molecular dynamics simulation of a membrane-peptide complex using
GROMACS and CHARMM-GUI with coarse-grained MARTINI force field

Free energy landscape

Ultra
coarse-grained
model

ails
det
Coarse-grained model

All-atom model
cale
me s
, t i
size
stem
sy
 Today job
construct and simulate this system using coarse-grained force field
The coarse-grained MARTINI force field
The lipid and peptide of our system
We use the CHARMM-GUI server to
construct the system and input files

www.charmm-gui.org
Then choose: membrane builder ==> bilayer builder
 Upper layer

Peptide

Lower layer
Please download this file
Having obtained coordinator le, map le and topology le, you can
perform simulation with GROMACS as you did for ala3 peptide

Provided by
CHARMM-GUI

Remove some bad contacts

between atoms during the
construction of the system
fi
fi
fi
 Equilibration
Obtained from previous step CHARMM-GUI
(6 steps in total)

Relax the system to

equilibrium condition
 Production run

CHARMM-GUI
After equilibration

Real simulation
 Analyze the simulated trajectories

1. Visualize the trajectory

2. Check energy terms
3. Area per lipid
4. Density of the membrane and water
5. Lateral diffusion of lipids
6. Conformational change of the peptide
 Note

To start GROMACS, please type:

source /opt/sdv/gromacs-2018.2/bin/GMXRC

To run xmgrace: please type

source activate xmgrace-env

To visualize structure with VMD, please open file: system.top

Then write the path explicitly

#include "/ibpc/seurat/nguyen/STUDENT/COURSE/charmm-gui-1552386511/gromacs/toppar/martini_v2.2.itp"
#include "/ibpc/seurat/nguyen/STUDENT/COURSE/charmm-gui-1552386511/gromacs/toppar/martini_v2.0_lipids_all_201506.itp"
#include "/ibpc/seurat/nguyen/STUDENT/COURSE/charmm-gui-1552386511/gromacs/toppar/martini_v2.0_ions.itp"
#include "PROA_P.itp"

[ system ]
; name
Martini system

[ molecules ]
; name number
PROA_P 1
DOPC 100
W 619
NA 8
CL 12

https://www.drugdesign.org/chapters/molecular-dynamics/