CHM695
Module 4
Empirical Force-fields
(April 15)
Potential energy is driven by the complex
electronic structural changes
Can I mimic these potential energy
changes without doing quantum
mechanics?
QM
harmonic
Bonding Interactions: to
describe bonds
H2 molecule
N
U (R ) =?
Our model should be based
on some experimental/
theoretical knowledge about
the system
H2
H1
1
U (R1 , R2 ) = kr (R
2
R0 )
1
U (R1 , R2 ) = kr (R
2
R = |R1
1
=
2
R2 |
parameters
R0 )
reproduce experiments/theory
2
2 2
k = 4 = 4 c v
Expt/theory
Harmonic
Actual
U(R)
R0
No dissociation for harmonic potential
H2O molecule
OH bonds can be treated by harmonic bonds
H-H as harmonic
OR
H-O-H angle as harmonic
DOF=3x3
H1
1
H2
1
2
0
U (R1 , R2 , R3 ) = kOH ROH1 ROH +
2
1
2
0
kOH ROH2 ROH +
2
1
2
0
kHOH H1 O H2 HOH
2
3
H2O2 molecule
2 angles [1 type]
1 torsion [1 type]
O1
3 distances [2 types]
O2
H2
H1
Potential for Torsions: Two kinds
Improper Torsion/Out-of Plane
1
2
U () = k,I
2
OR
1
2
U () = kr,I h
2
these describe out-of-plane motion
(Proper) Torsion
N
X
Vn
U( ) =
[1 + cos(n
2
n=0
)]
What value of N and gamma, and
Vn are suitable to mimic ethane
potential energy surface?
1
2
0
U (R1 , R2 , R3 , R4 ) = kOH RO1H1 ROH +
2
1
2
0
kOH RO2H2 ROH +
2
1
2
0
kOO RO1O2 ROO +
2
1
2
0
kHOO H1 O1 O2 HOO
2
1
2
0
kHOO H2 O2 O1 HOO
2
0
N
X
VHOOH
[1 + cos(n
2
n=0
H1 O1 O2 H2
H1 O1 O2 H2 )]
Some Other Examples:
Bond Potentials
Bond
R0 ()
kr (kcal mol-1
-2)
Csp3-Csp3
1.523
317
Csp3-Csp2
1.497
317
Csp2=Csp2
1.337
690
Csp2=O
1.208
777
Csp3-Nsp3
1.438
367
C-N (amide)
1.345
719
Angle
Angle
0 (deg)
k(kcal mol-1
deg-2)
Csp3-Csp3-Csp3
109.47
0.0099
Csp3-Csp3-H
109.47
0.0079
H-Csp3-H
109.47
0.007
Csp3-Csp2-Csp3
117.2
0.0099
Non bonding
Interactions
Electrostatic Interactions
Point charges: due to
electronic redistribution
when a molecule is formed
q+ R q12
1
1 q1 q2
U (R12 ) =
40 R12
Charges are only
conceptual
Charges are not
observables
Electronic density is an
observable
Electrostatic Potential Derived Point Charges
(ESP Charges)
Can we get point charges on atoms (i.e. atomic
positions) such that we can reproduce the
electrostatic potential?
ESP charges
Scheraga & his group,
Journal of Physical Chemistry,
1992, 96, 10276
Point charges are such that the
electrostatic potential obtained
from a quantum mechanical
calculation is mimicked for the
selected points
Fitting procedure is used here
from quantum
mechanics
Npoints
R=
wi
0
i
calc 2
i
I qI / |RI - r |
FORCE FIELDS FOR PROTEIN SIMULATIONS
Fig. 2. Charge models for the Amber potentials. (Left) HFthe style of Amber ff 94. (Right) polarizable, extra-point charge
Amber ff02-EP with the atomic polarizabilities in A3 given in par
Dispersive Interactions
He
He
such interactions are due to instantaneous
dipole-dipole interactions (dynamic
electronic correlation)
Can be generally mimicked by LennardJones potential
attractive
U (R) = 4
12
repulsive
AB =
AB
A B
+
2
Bond
Angle
Torsion
Non-bonded
Overall potential form:
bonds
UMM (RN ) =
torsions
X1
angles
k(x
x0 ) 2 +
X1
k(
0 ) 2 +
2
2
X Vn
(1 + cos n!)+
(2
"
#
)
12
6
X
qI qJ
IJ
ij
4IJ
+
RIJ
RIJ
RIJ
IJ
non-bonded pairs
for the development of
multiscale models for
complex chemical systems
2013 Nobel Prize
Chemistry
2013
Martin Karplus
Michael Levitt
Arieh Warshel
Simulation Protocol
Initial
structure
from X-ray
(pdb.org)
Solvate
protein
with water
and ions
Equilibrate
the system
BUT what if we dont have the
PDB structure?
Simulate
till
adequate
timescale
approximate
can MD predict the native
structure?
native
Can MD tell us why nature
chooses a specific folded
structure of a protein and
how!?
Protein structure
prediction
Simulation length will depends on the kinetic
barrier going from the starting structure to the
equilibrium/stable structure of the protein
Structure of the protein is
mainly determined by the
backbone structure
Rotational barriers of backbone atoms, and also side
chains can determine the tertiary structure of the
protein [Ramachandran Map]
limited torsional angles
are only feasible due to
torsional barriers
H-bonds, solvent interaction, temperature etc.
also affects the torsional values
Very large number of torsions
difficult to
predict the preferred values of torsions
Rotational barriers can be several and high!
long timescale simulations required
native state
= 10-5 to 10 seconds
Time Scale Bottleneck
MD time step: < 1 fs
Number of MD integration steps required
for observing a protein folding is at least
0.1s
8
= 10 s
1 fs
If one MD step costs 1 second
computational time, total computing
time would be 108 seconds = 3.2 years!!
To complete in one week, one step has to
b e c o m p l e t e d i n 6x1 0 -3 s e c o n d s
(computational time)!
2637 (2009).
www.sciencemag.o
of cell viability at approximately 1 mM L-CDME,
Materials and Met
16. L. S. Levitt, J. Phys. 49, 696 (1975).
and studies in rats, performed to measure oxiFigs. S1 to S10
17. R. Carta, G. Tola, J. Chem. Eng. Data 41, 414 (1996).
Table S1
18. I. Weissbuch, L. Addadi, L. Leiserowitz, L. Leiserowitz,
dative stress in the brain cortex, demonstrated
Atomic-Level Characterization
o
References
Science 253, 637 (1991).
adverse effects at dosages of approximately 500
Movies S1 to S3
19. of
G. Clydesdale,
R. B.David
Hammond,
Roberts,
J.etPhys.
E.K. J.Shaw
al.
Atomic-Level
Characterization
the Structural
Dynamics
of Proteins
mg/kg (mass of rat)
per day (3335).
Although
Chem. B 107, 4826Science
(2003).
David
E.
Shaw
et
al.
330, 341 (2010);
the pharmacokinetics
of
L-CDME
are
not
well
7 May 2010; acce
20.
I.
Weissbuch,
M.
Lahav,
L.
Leiserowitz,
Cryst. Growth Des.
Science 330, 341 (2010);
DOI:
10.1126/science.1187409
known, on the basis
of 10.1126/science.1187409
typical daily urine vol10.1126/science.1
3, 125 (2003).
DOI:
Atomic-Level Characterization
of the Structural Dynamics of Proteins
ally distinct sta
This copy is for your personal,
non-comm
terized
as ba
energy landsc
Substantial
both experime
If you wish to distribute
this article to
copies for your1
techniques, in c
1,2
1 others, you can order high-quality
1
David
E.
Shaw,
*
Paul
Maragakis,
Kresten
Lindorff-Larsen,
Stefano
Piana,
colleagues, clients, or customers by clicking here.
and the ways th
1
1
1
1
1
If
you
wish
to
distribute
this
article
to
others
, you can or
Ron
O.
Dror,
Michael
P.
Eastwood,
Joseph
A.
Bank,
John
M.
Jumper,
John
K.
Salmon,
Permission to
republish
or
repurpose
articles
or
portions
of
articles
can
be
obtained
by
them.
1
1
colleagues, clients, or customers by clicking here. It has pr
Yibing
Shan,
Willy Wriggers
following
the guidelines
here.
ally characteriz
to at
republish or repurpose articles
or and
portion
states
to el
The following resources related to this article arePermission
available online
the protein
guidelines
Molecular
dynamics (MD)
simulations
are
widely following
used
study
motions
at an atomic
www.sciencemag.org
(this
information
is current
as ofto
February
21, 2014
): here.
anisms involve
level of detail, but they have been limited to time scales shorter than those of many biologically states.
Updated
information and
services,
including
high-resolution
figures,processes
can
be found
in
the online
The
following
resources
related
to this article are
availa
critical
conformational
changes.
We
examined
two
fundamental
in
protein
All-atom
m
version of this article at:
www.sciencemag.org
(this information
asdesig
of
dynamicsprotein
folding and conformational change
within the folded stateby
means of is current
http://www.sciencemag.org/content/330/6002/341.full.html
tions are
extremely
long
all-atom
MD can
simulations
conducted on a special-purpose machine. Equilibrium
view of the m
Supporting
Online
Material
be found at:
Updated
information
andevents
services,
high-resolut
simulations
of a WW protein domain captured multiple
folding
and unfolding
that including cules
http://www.sciencemag.org/content/suppl/2010/10/13/330.6002.341.DC1.html
(9), prod
version
ofsimulations
this articleofat:
consistently
followadditional
a well-defined
pathway;
separate
the proteins
the potential t
A list of selected
articles folding
on the Science
Web
sites
related
to this article
can be constituent
http://www.sciencemag.org/content/330/6002/341.full.html
found at:
substructures
shed light on possible determinants of this pathway. A 1-millisecond simulation
shots generate
http://www.sciencemag.org/content/330/6002/341.full.html#related
of the folded protein BPTI reveals a small number
of structurally
distinctMaterial
conformational
Supporting
Online
can bestates
found at: tational constr
This article
cites 47
articles, 14 of is
which
can than
be accessed
free:
http://www.sciencemag.org/content/suppl/2010/10/13/330.6
whose
reversible
interconversion
slower
local
relaxations
within those states by a factor
http://www.sciencemag.org/content/330/6002/341.full.html#ref-list-1
of more than 1000.
1
A list of Press;
selected
site
This article has been cited by 47 articles hosted by HighWire
see:additional articles on the Science
D. E.Web
Shaw Resea
found at:
http://www.sciencemag.org/content/330/6002/341.full.html#related-urls
10036, USA. 2Cen
folding (1), signal transduction (2), the catalytic matics, Columbia
any biological processes involve func- http://www.sciencemag.org/content/330/6002/341.full.html#
aded from www.sciencemag.org on February 21, 2014
This copy is for your personal, non-commercial use only.
Addressing the grand challenge as we speak...
D. E. Shaw & his Anton Machine
http://www.youtube.com/watch?v=PGqCeSjNuTY
NEWS OF THE WEEK
Custom job. By speeding calculations, this Anton supercomputer can run all-atom
simulations (inset) 100 times
longer than can general purpose supercomputers.
proteins. But Shaw says
he and his colleagues are
already making progress on
that. In addition to building 11 supercomputers
incorporating 512 customdesigned computer chip
cores, or nodesShaw
donated one to the National Resource for
1000x faster
relatively simple calculations involved in
From the Science
Policy Blog
A National Academies report on how U.S.
universities have managed intellectual
property in the wake of the 1980 Bayh-Dole
Act has concluded that things are pretty
much hunky-dory but that schools may be
trying too hard to cash in on discoveries.
Universities instead should aim to disseminate technology for the public good, which
may mean passing up a more lucrative
licensing deal. http://bit.ly/bayh-dole-update
Anton Vs others
Machine
computational time for 1 MD step (s)
Intel Xeon chips
~104
Anton
~20
Anton has a speed up of ~1000
How Anton Does it?
Chips are designed for fast computation
and look-up of pairwise interactions
Communications between the chips are
extremely fast
For those who want to read more about it, see:
http://www.cs.utah.edu/hpca08/papers/6A_1_Larson.pdf
