Intro MM PES PDF

Molecular Mechanics
and
Potential Energy Surfaces
An introduction
UWI-Mona Per-Ola Norrby University of Gothenburg

Feb. 2011 Department of Chemistry
Molecular Modeling
Calculating properties of a structure

• Energy
• Optimum geometry
• Conformational distribution
• Reactivity
• Polarity
• Interactions w. other structures
– Solvation
– Docking
– Packing
• NMR
…

Energy
The basic property
The energy is a function of

the geometry, that is,
the coordinates (r).
F
E
Each system strives to minimize
the energy. Stable structures are
located at energy minima.
r
The force which acts on each atom

to bring it to the minima is the slope
of the energy function: F = -∂ E/∂ r
The most common task is to move all atoms in the direction of the force to
bring it to the minimum, where all other properties are calculated

Molecular Energy
Total energy: separate all electrons and nuclei completely
Atomization energy, heat of formation: separate into component atoms
Bond energy: separate molecule into two fragments.
Potential energy: relative to an arbitrary state defined as zero
— Steric energy: relative to a hypothetical unstrained conformation.

— Conformational energy: relative to lowest energy conformation

Calculating Energies
Target accuracy: 2-4 kJ/mol
Quantum Mechanics Hˆ " = E "

Solving the Schrödinger equation
The wavefunction: the position of every electron
Any property available from the wavefunction
!
Tradeoff between accuracy and speed
Currently ca 100 atoms with acceptable accuracy
Total energies: E = 0 at total separation
Molecular Mechanics
Simple, analytic functions, atomic positions only 2
E = # k (l " l0 )
Known systems only
Very fast, millions of atoms possible l
Steric energies: E = 0 for “unstrained” structures.
Only conformational energies valid !

Bond distortion energy
QM: E=0
Harmonic
Morse
MM: E=0

Torsional energy
gauche gauche
R anti R
R R R
R
2 kJ/mol 2 kJ/mol
60 180 300

Molecular mechanics
Molecular mechanics uses empirical functions to calculate the molecular energy

as contributions from distorted bonds, angles, etc.
E = ∑Es + ∑Eb + ∑Et + ∑Enb ( + ∑Ecross )

Simple example force field
Es = ks(l–l0)2
Eb = kb(θ–θ0)2 ω l
Et = ∑vncosnω θ
Enb = qiqj /εr + Ar-12 – Br-6
r
Ecross = ksb(l–l0)(θ–θ0)
Molecular mechanics
Bonds E
Harmonic Quartic
l Morse
Morse: Es = D[1–e–α (l–l0)]2

Harmonic: Es = ks(l–l0)2 (Hooke’s law)
Quartic: Es = ks(l–l0)2 [1+cs(l–l0)+qs(l–l0)2]
= ks∆l2 + kc∆l3 + kq∆l4

Molecular mechanics
Bond parameters
E = D[1–e–α (l–l0)]2 ≈ k(l–l0)2
E
H H
H
C C
1 H
" H H
k
!
l0 = 1.54 Å
rC–C

Molecular mechanics
Type-dependent parameters
H H
H
C C Carbon types:
H C H sp3
H
– small ring
sp2
E – carbonyl
– aromatic
– small ring
sp
k = 7.5 k = 6.3
100 atom types:

100*100*2 =
20000 parameters
l0 = 1.33 Å l0 = 1.50 Å
rC–C

Molecular mechanics
Nonbonded terms
Enb = Eelec + Evdw r
A/r12 or Ae–!r
Evdw = Ar-12 – Br-6
(Lennard-Jones)
Evdw = Ae–αr – Br-6
(Buckingham)
–B/r6
Harmonic force field:
only Lennard-Jones
Sum of vdW radii
Molecular mechanics
Electrostatics
Enb = Eelec + Evdw µi

qi r µj
Eelec = qiqj /εr r

qj
Eelec = qiqj /εr2
Eelec = (cosχ–3cosαicosαj)µiµj /εr3
Polarizable force field: either q or µ variable, environment-dependent

Molecular mechanics
Angles
θ
Harmonic: Eb = kb(θ–θ0)2
Sextic: Eb = kb(θ–θ0)2 +…+sb(θ–θ0)6
SHAPES: Eb = kb[1+cos(nθ +ψ)]
Alternative: Use no angle energy,

Harmonic but instead nonbonded interactions.
The POS model
180° Cusp
L
L L
M
L L
Coordination L
90° 180°
60 90 120 150 180

Torsional energy
gauche gauche
R anti R
R R R
R
4 kJ/mol 4 kJ/mol
60 180 300

Molecular mechanics
Torsions
ω
Et = v cos(nω + ψ)
Et = ∑vncosnω
Et = v1(1+cosω) + v2(1–cos2ω) + v3(1+cos3ω)
vdW
25 kJ/mol
+ v1
20 + v2
15 Real +
v3
10
=
5
vdW
only
0
0 60 120 180 240 300 360
0 60 120 180 240 300 360

Molecular mechanics Et ∝ cos3ω
Torsions: Why?
Force field interaction

Et ∝ cos2ω
Dispersion attraction
Empty !*
!E‡ Filled !
‡
Hyperconjugation

Molecular mechanics
Planarity & inversion
N L L
O N M
L L
Planes: Central atom & two substituents

Three substituents
Least-squares, all substituents ( & central )
Deviation: Distance to plane, other atom(s)

Angle between bond and plane
Dihedral angle (improper torsion)
EOOP = kOOP (dev) or kOOP (dev)2 or kOOP cos(dev)

Molecular mechanics
Planarity: why?
O
O O
! = 135°
! ! !0 = 120° ! = 150°
O
O
! = 135° ! = 120°
With planarity Without planarity

Molecular mechanics
Simple force field, harmonic, diagonal (nondiagonal)
E = ∑Es + ∑Eb + ∑Et + ∑Enb ( + ∑Ecross )

Es = ks(l–l0)2
Eb = kb(θ–θ0)2
Et = ∑vncosnω ω l
Enb = qiqj /εr + Ar-12 – Br-6 θ

Ecross = ksb(l–l0)(θ–θ0)
r

Molecular mechanics
Cross terms
Stretch-bend: l
Ecross = ksb(l–l0)(θ–θ0) θ
Large deviations: Ecross < Es=(l-l0)2
Other examples:
Stretch-stretch Bonds to same atom

Bend-bend Same central atom
Bend-bend Common bond, not same central atom
Torsion-stretch Central bond (eclipse elongation)
Bend-torsion-bend Central torsion bond common
Stretch and bend linear in deviation, torsion uses cosine term.
Diagonal force field: no cross-terms.

Molecular mechanics
Available force fields
Force field Notes

MM2 Dipole electrostatics, fractional π-bond orders
MM3 As MM2, added cross terms etc (vibrations)
MMX PCModel, MM2 with additions for coordination complexes
MM2*, MM3* MacroModel, charge electrostatics, no π, but substructures
CFF Many cross terms, vibrations, automated parameterization
CVFF Originally diagonal, many patch-on cross terms
CHARMM Bio-oriented, diagonal, harmonic
GROMOS Bio-oriented, diagonal, harmonic
AMBER Bio-oriented, diagonal, harmonic, requires HF electrostatics
OPLS Nonbonded interactions, solvation
Tripos Sybyl and Alchemy, diagonal, harmonic
Dreiding Main group elements, rule based parameters
UFF All elements, rule based parameters
MMFF Based on high level ab initio, automated parameterization
…

Functional groups
Predictivity
Chemical space Chemical space
Carbonyls
Alkanes Alkanes
Alkenes
Alcohols
Unknown Interpolation?

Structures
Energy minima, conformations
Energy
!E‡ Transition state,

!E reaction rates
Minimum,
observable

Energy minimization
Finding structures
3
2
5 4
Calculate forces, (the negative gradient, –∂ E/∂ r), move in the direction of the force
— Steepest descent, the step size is proportional to the force, preselected constant
— May diverge or oscillate
Line search: take longer and longer steps until the energy increases again
When a minimum is bracketed, use successive 3-point interpolation

Energy minimization
Steepest descent with line search
Calculate the forces (∂ E/∂ r, relatively time consuming)

Move in the direction of the force, calculate the energy at the new point (fast)
Continue in the same direction until a minimum is bracketed, interpolate (fast)
Line search finished, start again by recalculating forces
— Problem: all steps perpendicular, slow zig-zag towards minimum

Energy minimization
Conjugate gradient
Start with a steepest descent step, move in the direction of the force,
proportional to the force. Calculate the force in the new point.
Correct the step size; is there residual force in the same direction still?
Remember the old search direction, mix it with the new search direction.
— The conjugate search direction moves faster towards the minimum.

Energy minimization
Newton-Raphson minimization
Calculate gradients, ∂ E/∂ r, and how fast the forces change, ∂ 2E/∂ r 2 (the curvature)
Change the coordinate until the force is zero (=minimum), ∆r = –(∂ E/∂ r)/(∂ 2E/∂ r 2)
— Works well when the curvature is positive (close to a minimum)
— Moves uphill if the curvature is negative!
— Diverges if the curvature is close to zero
Use with line search to ensure downhill movement
Extremely efficient close to minimum, also in many dimensions (∂ 2E/∂ x∂ y, the Hessian)
Energy minimization
Truncated Newton-Raphson, Hessian update
Calculation of the Hessian (∂ 2E/∂ x∂ y) is extremely time-consuming

— Guess or estimate the Hessian, ensuring positive curvature (fast).
Calculate the force (–∂ E/∂ r), take a step using the guestimated Hessian.
With a perfect Hessian, the forces are now zero (=minimum).
Calculate the real forces, correct the Hessian.
Repeat with the new, improved Hessian (TNCG or BFGS).

Multiple minima
Conformations
The structure you find depend on where you start.
Free torsions (N) Conformations (3N)

1 3
2 9
3 27
5 243
10 59,049
15 14,348,907
Conformational searching
Generate new trial conformation, energy minimize, compare to previous conformations.
Retain if unlike, and within cut-off (typically 20-50 kJ/mol from lowest energy found).
Method Description
Cartesian stochastic search Move all atoms randomly
Monte Carlo search Rotate torsions randomly
Poling Add an energy penalty for similarity

to stored conformations
Molecular Dynamics (MD) Calculate velocities from the forces, evolve in time
Low Mode search Calculate vibrational spectrum, move along soft modes
Note: The terms “Monte Carlo” and “stochastic” just mean “random”. Each of these can also
be associated with other methods.

Cartesian stochastic search
Move all atoms randomly

Minimize energy
Generates high energy, distorted structures.

Best with simple force fields.
Can generate very strange local minima
C
H H
H H

Monte Carlo search
Bonds and angles unaffected,

low energy, easy to minimize
Rotate torsions (typically 120°),
minimize energy. 3N possible rotations.
MCMM (random), SPMC (systematic)
Ring breaking or flipping algorithm needed

Poling
r O
Add an energy penalty for similarity to stored conformations
— Similarity can be based on interatomic distances, penalty = ∑e –∆r2 N
2
1
1
Direct minimization on the new energy surface

will move away from any existing minimum.
2 Can be used in conjunction with other methods
1 Filling out minima, new, flat PES
3 Very costly after a few hundred steps
— Only compare selected measurements

Molecular dynamics
Study how a system evolves. Averages over time correspond directly to

experimental observables. Good for solvated calculations.
Start: Calculate forces on all atoms, F = ∂ E/∂ x, random velocities.

Using the atomic masses, calculate accelerations: F = ma
Take a small time step, ∆t. Calculate new velocities: vt+∆t = vt + a∆t
Using the velocities (possibly time-average velocities), calculate new positions for all
atoms. Adjust the velocities to a given temperature.
Calculate any property for time averaging. This could be energies, but also, for
example, NMR properties.
Repeat from “Start” until convergence is reached.

Molecular Dynamics (MD)
H
H H H H
H O H H O H O
O H O H
O H O H H O H
H
O H H
H H H O H H
H O
O H O H O
H H
Slow to cross barriers.
Can generate free
energies (∆G) directly.
O H
H H
H O O H
H H H H O
H H O
H O H
O H O
H
H
H
H O
H H
O O H
H H O O H
H
H O H
H
Dynamics
Why?
Calculate velocities and move all atoms, realistic

model. Other models basically at 0 K
Equilibrate position of hundreds of solvent

molecules
Find representative sample of conformations

(when barriers are low)
Make “chemical changes” slowly, system always

at equilibrium: free energy perturbation (FEP) or
thermodynamic integration (TI)

Low Mode search
Calculate vibrations (IR, from ∂ 2E/∂ x 2), move along soft modes
Slow to cross high barriers. Good for exhaustive search of local space
Coverage similar to MD, but more efficient, only moves towards low barriers

Population distributions
Boltzmann factor
T=300K (27°C)
p1=12%
p2=27%
p0=61%
!E1=4 kJ/mol !E2=2 kJ/mol
–!E0/RT
p1
p0 = e –!E1/RT
p0 = e
–!E0/RT –!E1/RT –!E2/RT
e +e +e
(cf. ∆G = –RTlnK)
At 25°C: 4 kJ/mol
2/1, 2 kJ/mol At -78°C, 12/1
Rules of thumb for ratios:
5/1, 4 kJ/mol At 25°C, 5/1
(also valid for relative rates)
10/1, 6 kJ/mol At 75°C, 4/1
(1.4 kcal/mol) At 250°C, 5/2

Optimizing to transition states
Maximize in one direction, minimize in all others. Which?
E ⇒ TS
Newton-Raphson converges to closest ⇒ Minimum ⇒ Minimum
stationary point (slope=force=zero).
"2E
2
<0
"2E "r "2E
2
>0 2
>0
"r "r
r
!
TS
! Newton-Raphson convergence
! radii
Requires accurate Hessian and/or special

TS updating algorithm

Coordinate driving
TS
TS
Fix coordinate (torsion) at different values

(–60° ⇒ 60°), optimize all other coordinates

Synchronous transit
Interpolation, requires geometries of

reactant and product.
Define the search direction as a line

connecting reactant and product. From
the chosen starting point, go uphill
along the search direction, and
downhill in all other directions.

Transition state energies and rates
""E ‡
Product A Reactant Product B

! "E A‡ "EB‡
"E A!
"EB!
! ""E ! !
!
!RT
!
Product ratio at equilibrium: r = e"##E
!
kBT #$E A‡ RT
Rate constant, formation of A: kA = " e
h
! "##E ‡ RT
Kinetic ratio, A/B: k A kB = e
UWI-Mona
!
Per-Ola Norrby University of Gothenburg
!
Molecular Vibrations
and
Molecular Energies

Molecular vibrations
All molecules vibrate, always. From the atomic masses and the energy curvature,
the frequencies can be calculated. Depending on temperature and curvature, higher
vibrational states might be populated, especially for soft modes.

Anharmonicity
It is easier to elongate than to compress a bond. The potential energy well gets
broader than the harmonic model predicts, so energy levels are closer together.
re Boltzmann average,
gives rg, rz, …
Measured bond lengths are vibrational averages, generally longer than

calculated bond lengths.

Vibrational energy
At 0 K, only the lowest vibrational level is populated. The energy difference between
this level and the bottom of the well (the calculated potential energy, E0) is the “Zero
point energy”, ZPE, which can be calculated from the vibrational frequencies: ZPE =
∑hν/2
UT (!HT)
ZPE
E0
At higher temperatures, some particles are in a higher vibrational state. The average
energy of all particles gives the macroscopic “inner energy”, U. At constant volume
and pressure, this is the same as the more useful enthalpy, H. At 0 K, U0 = H0 = E0 +
ZPE.
Most computational packages will calculate ZPE, U, H, S, and G if you request a
vibrational analysis.
Degenerate conformations
Conformational entropy
In the example below, there is one global minimum and two equivalent
(experimentally indistinguishable) higher conformations.
gauche gauche
R anti R
R R R
R
2 kJ/mol 2 kJ/mol
60 180 300
From the “rules of thumb”, at room temperature the ratios are 1:2:1.
That means that there are equal amounts of gauche and anti, that is, the total free
energy of gauche and anti are equal. The free energy of gauche is lowered by a
conformational entropy contribution equal to Rln2.
– The entropy contribution is always RlnW, where W is the number of states.

Entropy and free energy
∆G = ∆H – T∆S
The entropy, ∆S = RlnW, just comes from the fact that there are many possible states
in each conformation. ∆G could be obtained equivalently by summing the energy
over all possible states, but it is easier to calculate an energy minimum once, and then
get the entropy for that minimum by counting the number of equivalent
conformations, the population in all vibrational states, etc.
Higher energy, but more possible states, therefore

higher population and thus lower free energy
Gas phase free energy
Gtot = E0 + ZPE + Hvib + T Svib + Grot/trans + T Sconf
Temperature independent:
E0 The potential energy at the calculated minimum

ZPE Zero point energy
Temperature dependent (no contribution at 0 K):
Hvib From averaging over vibrational levels

(Note: many packages include ZPE in this term!)
Svib Vibrational entropy
Grot/trans Contribution from translation and rotation
Rule of thumb:
30±10 kJ/mol for each molecule at room temperature
Sconf Conformational entropy. Either R ln(number of conf.)
or sum over all conformations, including degenerate ones.

Solvent
Charge stabilization
Charges, especially localized anions, are much more stable in solvent.
F Extremely strong base in gas phase, less so in solvent

O
Delocalized charge, less difference between gas phase and
solvent. About equal to fluoride in water.
O
Charge separation is almost impossible in gas phase
O O
gas phase
H3 N O water H2 N OH

Solvation Continuum: cavity with
Gas phase Born: Free energy gain from compensating surface charges
exposed surface of polar atoms !" !+ !+
O !"
!" !+
O
O O H3 N
H3 N !+
NH3 !"
!" O !+
O !" !+
!+ !+
Reduced charges O Explicit solvation, very realistic and

H3 N demanding
O
O O
Micro- H3 N H3 N
solvation
O O

Solvent vs. gas phase
Conformational collapse
Polar functional groups will attract each other strongly in gas phase. In solvent,
separation of charges may actually be favorable.
gas phase water

(bioactive)
If you search for conformations without considering the solvent, you may well miss
all the important ones!

Evaluating
and
Parameterizing
Force Fields

Evaluating Force Fields
Compare computed properties to reference data!
Structures: Compare bonds, angles. Avoid long distances

(crystal packing forces).
Energies: Experimental energies often free energies!

Use enthalpies, compensate for entropy, solvation.
Vibrations: Needed for accurate force constants.

(IR) Very hard to assign, unless from QM.
Charges: Not observable, use dipoles or QM charges.

Crystal structures
X-ray gives electron densities, not nuclear positions
— very different for hydrogen Nucleus
Neutron diffraction gives nuclear positions C H
Center of electron density

Other error sources:
• Crystal packing
• Libration, disorder
Comparison methods:
• Overlay
• Bond & angle list
• Crystal structure energy

Crystal packing
Tight packing is favored.
41°
Crystal structures are more “planar”
than ideal in vacuo structures
Only soft modes are distorted crystal gas

— Bonds and most angles are good
— Torsional angles can deviate Large distance error
H H H H
Small changes accumulate
Small angle error

Libration & disorder
Atomic positions from crystallography are average positions.

Thermal wagging motions (librations) or disorder may give
unphysical structures.
Libration of a bond
120° Disorder
Pd
Pd
Average Apparent
atom short bond
position Average structure
Pd Large angle, short bonds
Phenyl libration Apparent phenyl

Overlaying structures
2
Root mean square error: rms = # ( xobs " xcalc )
Summation is done over all cartesian coordinates

Rotate and translate structure
! until rms minimized (automatic)
Final rms value gives agreement between two structures
• Sensitive to soft mode errors

• Sensitive to accumulation of small errors
Not general as “goodness” indicator.

If used, select overlay atoms carefully!

Comparing structures
Compare lists of bonds and angles
Calculate rms for bonds and angles separately
• Insensitive to torsional errors

• No error accumulation
Inspect largest errors, find a rationale
• Inaccurate calculation or experimental error?

• Soft modes (some angles) and crystal packing

Crystal structure energies
The energy of a crystal structure, compared to the energy of an

optimized structure, can be used as a “goodness” measure.
Small bond length deviations give large energy penalties.
Structures should be relaxed

E
by constrained optimization.
Flat-bottom potential gives no
penalty for small deviations
Neutral structures should be within 10-20 kJ/mol of minimum

Comparing computed structure
QM and MM structures are both energy minima, can be directly

compared (zero K, in vacuo, no vibrational contributions)
Well-developed force fields

are usually at least as accurate
as standard QM methods. Use
re validation against QM results
with caution.

Molecular Mechanics Energies
Empirical force fields calculate distortion energies.
2 2 ' qi q j A B *
E= # ks (l " l0 ) + # kb ($ " $0 ) + # vn cos n% + # )
r
+ 12 " 6 ,
bonds angles torsions r-3 bonds ( & r r +
The energy from each term differs from “chemical” potential energy by a constant,
unknown term.
!
QM: E=0
Comparison of force field energies Harmonic
are only valid when all such unknown
terms cancel (i.e., comparing Morse
conformers)
∆Econf can be compared to

experimental conformational energies MM: E=0

Energies
Experimental energies are usually free energies obtained from
equilibrium constants.
∆G = RT lnK ∆G = ∆H – T∆S
Free energies differ from calculated potential energies
Gtot = E0 + ZPE + Hvib – T Svib + Grot/trans – T Sconf + Gsolv
Each contribution can be calculated before comparison.

Experimental energies
∆G = ∆H – T∆S = RTlnK ⇒ RlnK = ∆S – ∆H/T
RlnK
slope = –!H
x
!S x x
x x
x
1/T
Common assumption: ∆H ≈ ∆E
(∆ZPE ≈ 0, ∆∆S ≈ 0, ∆∆Gsolv ≈ 0)

Reference Energies
t-Bu
CH3 CH3
H H !H=4.1 H CH3 !G=4.4
t-Bu t-Bu
H H kJ/mol H H t-Bu kJ/mol
CH3 H
t-Bu t-Bu
R
H R' R'
O !H=2.9
H O -6 - 23
kJ/mol R
kJ/mol
F F H H
H H
F !H=3.3 H H
H H kJ/mol H H H
H F H H
!E‡=12.0 kJ/mol !E‡=7.5 kJ/mol
Force field comparison
Liljefors, Gundertofte, Norrby, Pettersson, in Computational Medicinal Chemistry and Drug Discovery,
Eds: Tollenaere, Winter, Langenaeker, Bultinck; Marcel Dekker: New York, 2003, 1-28
Mean Absolute Error (MAE), kJ/mol
14
12 Conjugated
Halocyclohexanes
10 Halides
Cyclohexanes
8 Nitrogen
Oxygen
6 Hydrocarbons
Rotation barriers
4
0
MM2*
MM3*
MM2(91)
MM3(92)
MMFF
Amber*
OPLS_AA
HF/6-31G*
CHARMm_23
PM3
CVFF
Dreiding2.21
CFF91
CFF99
B3LYP/6-31G*
Sybyl5.21
UFF1.1
Extending MM
Developing parameters
ω l
1) Nonbonded parameters
2) Reproduce structures
θ
3) Energies, vibrations, …
4) Properties
r
parameter
Es = ks(l–l0)2 Eb = kb(θ–θ0)2
observable
Et = vcos(nω +φ) Enb = qiqj /εr + Ar-12 – Br-6

parameter

Estimating parameters
1) From related parameters

Same type of parameter, similar atom types
2) From reference data

Ideal bond lengths and angles from unstrained structures,
force constants from comparing strained and unstrained structures,
charges from QM calculations, ...
3) Rule based
Bond lengths from sum of covalent radii, angles and torsions from
gross hybridization type, charges from electronegativity.

Parameters & observables
Es = ks(l–l0)2

Force field quality
The penalty function
Reference data y :
bond length l, angle θ, torsion ω, energy E, dipole µ, ...
Penalty function (error function)
χ2 = ∑wi2(ycalc –yobs)2
Optimal force field = minimal χ2

Force field quality
Weight factors
Dimensionless χ2 = ∑wi2(ycalc –yobs)2
Conversion and weighting
An error of 1° is less important than 1 Å !

Weight each data point by the inverse of the acceptable error
The rms error of the force field is acceptable when χ2 is lower
than the number of data points N.
Example: Data type Acceptable error Weight factor
bond length 0.01 Å 100

angle 1° 1
conformational energy 1 kJ/mol 1
QM charge 0.1 au 10

Parameterization
P.-O. Norrby T. Liljefors, J. Comput. Chem. 1998, 19, 1146-1166
Gather reference data Penalty function:
χ2 = ∑w 2(y-yref)2
Minimize χ2 by varying the parameters

Create MM model
• Grid search
• Monte Carlo
• Simplex
• Genetic Algorithm
Penalty function
Utilize derivatives, ∂ χ2 / ∂ p
• Steepest descent
• Conjugate gradient
Refine parameters • Newton-Raphson

Simplex
1) Initial Simplex 2) Reflection 3) Expansion or contraction
p2 p2 p2 expansion
best new best reflection
worst inversion worst

center contractions
p1 p1 p1
p1
4) Next cycle
8
9
p2 new
7
6
5 worst
p1
4
Biasing
3 p2
new move inversion
best center towards
2
1
best point
worst
p1
Least-squares optimization
The Newton-Raphson method
∂ χ2/∂ p
Newton-Raphson, 1 dimension: ∆p = –
∂ 2χ2/∂ p 2
Multidimensional: [∆pi] = –[∂ 2χ2/∂ p i∂ pj]–1[∂ χ2/∂ p i]
Approximate:
Gauss-Newton: [∂ yk/∂ p i]T[∂ yk/∂ p i] [∂ yk/∂ p i]T [∆yk]
Time-consuming, but good convergence

Summary, Molecular Mechanics
Fast method
Needs parameters
Structures
Conformational energies
Searching conformational space
Several methods
– Main division: small organics or biomolecules
Compare and validate


Intro MM PES PDF

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Intro MM PES PDF

Uploaded by

Copyright:

Available Formats

Molecular Mechanics

UWI-Mona Per-Ola Norrby University of Gothenburg

Calculating properties of a structure

UWI-Mona Per-Ola Norrby University of Gothenburg

The energy is a function of

The force which acts on each atom

UWI-Mona Per-Ola Norrby University of Gothenburg

Total energy: separate all electrons and nuclei completely

Atomization energy, heat of formation: separate into component atoms

Bond energy: separate molecule into two fragments.

Potential energy: relative to an arbitrary state defined as zero

— Steric energy: relative to a hypothetical unstrained conformation.

UWI-Mona Per-Ola Norrby University of Gothenburg

Quantum Mechanics Hˆ " = E "

UWI-Mona Per-Ola Norrby University of Gothenburg

UWI-Mona Per-Ola Norrby University of Gothenburg

UWI-Mona Per-Ola Norrby University of Gothenburg

Molecular mechanics uses empirical functions to calculate the molecular energy

E = ∑Es + ∑Eb + ∑Et + ∑Enb ( + ∑Ecross )

Morse: Es = D[1–e–α (l–l0)]2

UWI-Mona Per-Ola Norrby University of Gothenburg

E = D[1–e–α (l–l0)]2 ≈ k(l–l0)2

UWI-Mona Per-Ola Norrby University of Gothenburg

100 atom types:

UWI-Mona Per-Ola Norrby University of Gothenburg

Enb = Eelec + Evdw r

Enb = Eelec + Evdw µi

Eelec = qiqj /εr r

Polarizable force field: either q or µ variable, environment-dependent

UWI-Mona Per-Ola Norrby University of Gothenburg

Alternative: Use no angle energy,

UWI-Mona Per-Ola Norrby University of Gothenburg

UWI-Mona Per-Ola Norrby University of Gothenburg

UWI-Mona Per-Ola Norrby University of Gothenburg

Force field interaction

UWI-Mona Per-Ola Norrby University of Gothenburg

Planes: Central atom & two substituents

Deviation: Distance to plane, other atom(s)

EOOP = kOOP (dev) or kOOP (dev)2 or kOOP cos(dev)

UWI-Mona Per-Ola Norrby University of Gothenburg

With planarity Without planarity

UWI-Mona Per-Ola Norrby University of Gothenburg

Simple force field, harmonic, diagonal (nondiagonal)

E = ∑Es + ∑Eb + ∑Et + ∑Enb ( + ∑Ecross )

Enb = qiqj /εr + Ar-12 – Br-6 θ

UWI-Mona Per-Ola Norrby University of Gothenburg

Stretch-stretch Bonds to same atom

Stretch and bend linear in deviation, torsion uses cosine term.

Diagonal force field: no cross-terms.

UWI-Mona Per-Ola Norrby University of Gothenburg

Force field Notes

UWI-Mona Per-Ola Norrby University of Gothenburg

Chemical space Chemical space

UWI-Mona Per-Ola Norrby University of Gothenburg

!E‡ Transition state,

UWI-Mona Per-Ola Norrby University of Gothenburg

UWI-Mona Per-Ola Norrby University of Gothenburg

Calculate the forces (∂ E/∂ r, relatively time consuming)

— Problem: all steps perpendicular, slow zig-zag towards minimum

UWI-Mona Per-Ola Norrby University of Gothenburg

UWI-Mona Per-Ola Norrby University of Gothenburg

Calculation of the Hessian (∂ 2E/∂ x∂ y) is extremely time-consuming