C7790 Introduction to Molecular Modelling -1Lesson
9
Model
C7790 Introduction to Molecular
Modelling
TSM Modelling Molecular Structures
Petr Kulhánek
kulhanek@chemi.muni.cz
National Centre for Biomolecular Research, Faculty of Science
Masaryk University, Kamenice 5, CZ-62500 Brno
PS/2020 Distant Form of Teaching: Rev1
C7790 Introduction to Molecular Modelling -2Reality
vs Simulation
Problem
Model Result
approximations
Result
Result
experiment
approximation
prediction
Why?
⚫ incomplete theory
⚫ insufficient performance of current and future (?) computers
Solution ...
⚫ use approximation for solution of problems using the available computing
capacity
Is it possible to accurately simulate the reality around us?
Unfortunately, no :-(
Model chemistry
(Calculation method)
C7790 Introduction to Molecular Modelling -3Do
we need a model?
)()(ˆ rr kkk EH  =
In theory, no model is needed because it is outcome of SR equation solution.
The only input: nuclei and electrons and their description of interactions and motions
Solutions: energy and wavefunction describing QM states
(microstates)
In practice, we need to employ the BO approximation, which then requires a model (R).
C7790 Introduction to Molecular Modelling -4Born-Oppenheimer
Approximation
),(),(ˆ RrRr  EH =
෡𝐻𝑒Ψ(𝐫, 𝐑) = 𝐸𝑒(𝐑)Ψ(𝐫, 𝐑) ෡𝐻 𝑅 𝜒(𝐑) = 𝐸 𝑉𝑅𝑇 𝜒(𝐑)
electronic properties of the molecule
vibrational, rotational, translational
motions of a molecule
)(),(),( RRrRr  =
Born-Oppenheimer approximation
✓
We need a structure (model) to calculate
energies.
C7790 Introduction to Molecular Modelling -5What
is a model?
A model is smallest representation of studied system, which can describe studied
phenomena by chosen computational method (model chemistry).
molecule (molecules)
in vacuum
molecule (molecules) in
implicit environment
molecule (molecules) in
explicit environment
is explicitly modelled by
atoms/molecules
environment (typically solvent, membrane, etc.)
is implicitly modelled as a mean
field environment representation
(e.g., polarizable dielectric)
Main model types:
C7790 Introduction to Molecular Modelling -6Is
environment important?
In molecular modelling, it is not good idea to neglect environment even when qualitative
outcome is required.
! Neglecting environment can lead to wrong conclusions !
Modelling of molecules in vacuum (only) must be carefully justified.
bambus[6]uril/anion interaction
vacuum binding affinities:
F- > Cl- > Br- > Isolvent
(MeOH/CH3Cl) binding affinities:
I- > Br- > Cl- > Fthe
order is changed due to anion
desolvation energies
SLAVÍK, Jan. Počítačové modelování glykolurilových struktur, Bachelor's thesis. Masaryk
University, Brno, 2010
C7790 Introduction to Molecular Modelling -7Is
environment important? cont.
molecule (molecules) in
implicit environment
molecule (molecules) in
explicit environment
➢ Mainly used in QM calculations
▪ PCM (Polarizable continuum model)
▪ COSMO (Conductor like screening
model)
▪ …
➢ but also in MM
• PB (Poisson–Boltzmann solvent
model)
• GB (Generalized Born solvent model)
• 3D-RISM (3D reference interaction site
model)
• …
➢ Too complex for QM calculations (rarely
used).
➢ Typically used in MM and MD
• TIP3P (water model)
• SPC/E (water model)
• …
Homework:
How accurately different solvent
types (implicit/explicit) can describe
interactions at solute/solvent
interface?
C7790 Introduction to Molecular Modelling -8Model
is a compromise
Problem
Model Result
approximations
Result
Result
experiment
approximation
prediction
Model chemistry
(Calculation method)
too big
too complex
We need to choose between accuracy (computational feasibility) and model reliability
(reasonable representation of studied system).
C7790 Introduction to Molecular Modelling -9Model
sizes and time scales
Hsu, C.C., Buehler, M.J. & Tarakanova, A. The Order-Disorder Continuum: Linking Predictions of Protein
Structure and Disorder through Molecular Simulation. Sci Rep 10, 2068 (2020).
https://doi.org/10.1038/s41598-020-58868-w
C7790 Introduction to Molecular Modelling -10-
Example
➢ DNA (15-nt long)
➢ 948 atoms
➢ c(DNA)=7 mM
➢ explicit ions
➢ n(Na+)=35, c(Na+)=244 mM
➢ n(Cl-)=7, c(Cl-)=49 mM
➢ effective c(NaCl)=154 mM*
➢ explicit water (TIP3P model)
➢ n(H20)=7592
➢ 22776 atoms
➢ PBC with truncated octahedral box
➢ largest subscribed sphere Rin=29 Å
*Machado,M.R. and Pantano,S. (2020) Split the Charge Difference in Two! A Rule of Thumb for
Adding Proper Amounts of Ions in MD Simulations. J. Chem. Theory Comput., 16, 1367–1372
C7790 Introduction to Molecular Modelling -11Where
to get a model?
➢ In silico modelling
➢ small molecules
➢ 2D -> 3D conversions (high-throughput modelling, virtual screening)
➢ ab initio prediction of biomolecular structures
➢ Modelling based on experimental structures
➢ small molecules
➢ large molecules (proteins, DNA, biomolecular complexes, …)
➢ Experimentally guided modelling
➢ NMR (NOE contacts, …)
➢ cryoEM, SAXS (electron density, shape, …)
➢ Similarity modelling
➢ in silico modification of experimental structures
➢ homology modelling
C7790 Introduction to Molecular Modelling -12In
silico modelling
Avogadro Nemesis
Other software (commercial):
➢ Spartan, Hyperchem
➢ SCM (ADF)
➢ …
free drawing of
molecular structures
piecewise assembly of
molecular structures
For modelling, we need 3D structures.
)(RE
https://en.wikipedia.org/wiki/Comparison_of_software_for_molecular_mechanics_modeling
Overview of software:
C7790 Introduction to Molecular Modelling -13-
2D vs 3D structure
2D structure contains information about the
atoms and bonds. This information
describes the constitution (topology) of the
system.
3D structure contains information on the
spatial distribution of atoms in space. Other
information (e.g., bonds) is computable.
3D2D
benzoic acid
✓
suitable for modelling
C7790 Introduction to Molecular Modelling -14-
3D <-> 2D conversions
conversion is easy
3D2D
benzoic acid
conversion is difficult or impossible
C7790 Introduction to Molecular Modelling -15-
3D/2D conversions, complications
cyclohexane
chair conformation
twist boat conformation
For small molecules, 2D->3D conversion is
possible. Usually, the most stable
conformer is modelled.
✓
C7790 Introduction to Molecular Modelling -16Most
common formats:
➢ SMILES (Simplified Molecular-Input Line-Entry System)
➢ InChI (IUPAC International Chemical Identifier)
➢ InChIKey (IUPAC International Chemical Identifier Key)
2D structure usage
Representation of molecules in 2D formats is employed mainly for:
• storing information in databases
• searching in such databases (InChiKey and other variants)
• predicting the chemical properties of molecules using chemoinformatic approaches
(machine learning)
• automatic structure generation, generating libraries of molecules (computer aided
combinatorial chemistry) - virtual screening
benzoic acid
C(=O)(O)c1ccccc1
InChI=1S/C7H6O2/c8-7(9)6-4-2-1-3-5-6/h1-5H,(H,8,9)
WPYMKLBDIGXBTP-UHFFFAOYSA-N
hash
constant length, possible collisions
C7790 Introduction to Molecular Modelling -17Virtual
screening (motivation)
Bacteria
Early inhibition of bacterial lectin surface hinder bacterial adhesion to host cells.
Potent inhibitor (glycomimetics) can be used in treatment of bacterial infections.
(development of new antibiotics)
Cell
Bacteria
Cell
saturation with potent inhibitor
protection
C7790 Introduction to Molecular Modelling -18Virtual
screening
Docking
• method that tries to find
geometry of ligand/receptor
complex
receptor
ligands (ligand library)
Virtual screening
• identification of compounds with
highest affinity towards receptor
• plus special properties …
Which of them is the best?
C7790 Introduction to Molecular Modelling -19Screening
library
How to obtain the library?
Potential ligand sources:
• in silico modelling (2D -> 3D conversion)
• precalculated/experimental structure libraries
• PubChem (https://pubchem.ncbi.nlm.nih.gov/)
• ZINC (https://zinc.docking.org/)
C7790 Introduction to Molecular Modelling -20-
3D/2D conversions, biomolecules
closed form of the enzyme
open form of the enzyme
The same primary structure
(amino acid sequence).
different
conformations
In molecular modelling, ab initio modelling
(prediction) of biomolecular structures is
challenging task.
C7790 Introduction to Molecular Modelling -21Experimental
structures
It contains about half a million structures of small molecules determined by Xray
and neutron diffraction. Software for working with data:Mercury
http: // www.ccdc.cam.ac.uk/Solutions/CSDSystem/Pages/Mercury.aspx
Cambridge Structural Database (CSD)
http://www.ccdc.cam.ac.uk/Solutions/CSDSystem/Pages/CSD.aspx
It contains about 94 thousand structures biomolecular systems determined
mainly by X-ray structural analysis.
Protein Data Bank (PDB)
http: //www.pdb.org
Experimental method Proteins (P) Nucleic acids (NA) P / NA complexes Other Overall
X ray 77445 1481 4069 3 82998
NMR 8851 1046 193 7 10097
electron microscopy 469 45 129 0 643
status in September 2013
C7790 Introduction to Molecular Modelling -22Experimental
structures, cont.
➢ Experimental structures are usually sources for models of biomolecular structures or
complicated small molecules.
➢ Due to low resolution and molecular flexibility, some parts might be unresolved.
➢ Missing parts need to be modelled in silico
➢ hydrogen atoms (assignment can be sensitive to pH, PROPKA,
https://github.com/jensengroup/propka)
➢ flexible protein loops (Modeller, https://salilab.org/modeller/)
➢ Structures can be influenced by the crystal packing.
➢ It is advisable to check source electron density, especially for low-resolution structures.
➢ Check B-factors to evaluate structure quality.
C7790 Introduction to Molecular Modelling -23-
B-factors
The Debye–Waller factor (DWF, B-factor, temperature factor) is used in condensed matter
physics to describe the attenuation of X-ray scattering or coherent neutron scattering
caused by thermal motion.
For protein structures:
The B-factors (B) can be taken as indicating the relative vibrational motion of different
parts of the structure.
ORTEP diagram drawn with 40% ellipsoid
probability for non-H atoms
➢ Atoms with low B-factors belong to a part of the structure that is well ordered.
➢ Atoms with large B-factors generally belong to part of the structure that is very flexible.
𝐵 =
8𝜋2
3
< 𝑢2 >
u - displacement of scattering center (atoms)
<> - time or thermal average
C7790 Introduction to Molecular Modelling -24Similarity
modelling
http://www.unil.ch/pmf/en/home/menuinst/technologies/homology-modeling.html
Homology modelingModifying existing experimental structures
Available experimental structure(s) is (are)
modified
• structure substitution
• assembly of complexes
• …
C7790 Introduction to Molecular Modelling -25-
Summary
➢ Proper model is a key element for molecular modelling.
➢ Any error in the model propagates to calculated results.
➢ Therefore, it is worth to spent some time to check validity of the model (especially for in
silico modelled parts).
➢ It is also advisable to put some effort in cleaning/improving the model (atom names,
etc.) as it can save a lot of time in later analyses.