MM and MD, multi-scale (coarsed grain approaches)

1

• MM and MD, multi-scale (coarsed grain approaches)
  
• QM state-of-the art? an overview
• Andreas Koester, deMon Developers
• 3. ReaxFF
• 4. Kinetic Monte Carlo a glimpse
• 5. Genetic regulation a start stochastic
  networks, DFT, ReaxFF
• Rui Zhu, Andre Ribeiro, Stuart Kauffman, Adri
  van Duin, Bill Goddard
• Van der Waals interactions and DFT Some thoughts
• Petr Jurecka, Pavel Hobza, Jiri Cerny,
  Yue Zhang, Alberto Vela

2

• Complex systems
• Break them up
• Use many techniques
• MULTI-SCALE
• Make the techniques work together INTEGRATE!

From Molecules to Cells
QM
QM
MM
etc.
Q-chem
Stat-mech
dynamics
FE,CFD,
3
Molecular Mechanics, Molecular Dynamics state
of the art?

• F ma
• F - dE/dr
• E Vbonds Vangles Velectrostatic Vvan der
  Waals VH-bonds V polarization (?) .
• Drive MD with forces from empirical E
• Multi-scale coarse grained approaches allow large
  systems for long times

4
The multiscale challenge for biomolecular
systemscoarse-grained modelingJ.-W. CHU, S.
IZVEKO and G. A. VOTHMolecular Simulation, Vol.
32, Nos. 34, 15 MarchApril 2006, 211218
Multi-scale coarse-graining extract CG
parameters from underlying fine-grained
interactions
In the FM method, the set of M parameters, (g1, .
. . ,gM), are determined by minimizing the
average error in forces over the whole
configuration data in an atomistic molecular
dynamics (MD) simulation
Use a short atomistic simulation to fix the
parameters for a long CG simulation
5
The multiscale challenge for biomolecular
systemscoarse-grained modelingJ.-W. CHU, S.
IZVEKO and G. A. VOTHMolecular Simulation, Vol.
32, Nos. 34, 15 MarchApril 2006, 211218
Application of force-matching to
simulate dimyristoylphosphatidylcholine (DMPC)
lipid bilayer
Fine grain MD 64 DMPC, 1312 water, equilibrate
6ps, time step 2fs, 400ps simulation, save every
0.1ps, i.e. 4000 configs.
Coarse grain MD Same system size, 50 times
faster (more for larger systems)
6
Multiscale modeling of lipids and lipid
bilayersAlexander P. LyubartsevEur Biophys J
(2005) 35 5361
Formation of bicelles and vesicles
Vesicle formation starting from a square membrane
sheet of 3,522 lipids. A cross-section is shown
at different time moments (nonscaled time is
given in ns)
7
Assembly of lipoprotein particles revealed by
coarse-grained molecular dynamics simulationsAmy
Y. Shih, Peter L. Freddolino, Anton Arkhipov ,
Klaus Schulten Journal of Structural Biology 157
(2007) 579592
Good Cholesterol
36 018 CG beads 2 half-circle proteins 160 DPPC
lipids Represents about 500 000 atoms??
25fs time step
8
Klaus Schulten Quotes (non-verbatim, from my
notes)Boston Amer. Chem. Soc. Mtg, August 2007

• Computational microscope (NAMD VMD)
• 3 million atoms, 100 ns
• With a petaflop, simulate true life
• Vescicle 30nm diameter, 50 million atoms
• Photosynthetic chromatophore, 100 million atoms
• Bacterial flagellum, a billion atoms
• VMD a tool to think
• Describe entire living cells at atomic resolution
  

9
Klaus Schulten David Shaw
10
Cell and biomolecular mechanics in silicoAshkan
Vaziri and Arvind Gopinathnature materials VOL
7 JANUARY 2008 p. 15
Network models based on Monte-Carlo and molecular
dynamics
But
11

• Theres no chemistry yet

12
KS-DFT Choices

• Exc, vxc d Exc / dr

• Gaussians
• Slaters
• Numerical Functions
• Plane waves
• APW etc.

• Exact
• Fitting
• Multipoles

• PP
• ECP
• MCP

• L(S)DA
• GGA
• Hybrid
• AC
• LAP,TAU
• (meta-GGA)
• OEP

13
DFT publications 1964- 2005
1998 Nobel prize to Walter Kohn
14
deMon developersGaussian-based KS-DFT other
toolshttp//www.demon-software.com

• deMon-KS (molecules, clusters)
• (pre-deMon Brett Dunlap, Jan Andzelm, René
  Fournier), Alain St-Amant, Annick Goursot, Lars
  Pettersson, Emil Proynov, Mark Casida, Fiona Sim,
  Nathalie Godbout, Jingang Guan, Nino Russo,
  Claude Daul, Andreas Köster,Helio Duarte,
  Christine Jamorski, Vladimir Malkin, Olga
  Malkina, Alberto Vela, Klaus Hermann
• Bloch-deMon (solids, slabs, surfaces)
• Hisayoshi Kobayashi, Emil Proynov
• deMon-MD, deMon-MC (BOMD, QM/MM, QM/MM)
• Dongqing Wei, Thomas Heine, Annick Goursot
• deMon-properties
• Martin Leboeuf, Andreas Köster

15
deMon developersGaussian-based KS-DFT other
toolshttp//www.demon-software.com

• deMon-NMR (chem. shifts, J-couplings, hyperfine,
  Mössbauer)
• Vladimir Malkin, Olga Malkina,Mark Casida, Elisa
  Fadda, Seongho Moon
• deMon-DynaRho (time-dep. DFT, spectra)
• Mark Casida, Christine Jamorski, Jingang Guan,
  Sébastien Hamel, Elisa Fadda
• deMon-Functionals
• Emil Proynov, Alberto Vela, Henry Chermette,
  Petr Jurecka, Annick Goursot, Thomas Heine, Yue
  Zhang
• deMon-embed (embedded clusters)
• Helio Duarte, Seongho Moon
• deMon-Interact (distributed multipoles)
• Suzanne Sirois
• deMon-KSCED (constrained density embedding)
• Tomasz Wesolowski

16
deMon developersGaussian-based KS-DFT other
toolshttp//www.demon-software.com

• deMon-2k
• Andreas Köster, Gerald Geudtner, Annick Goursot,
  Thomas Heine, Alberto Vela, Ulises Reveles,
  Roberto Flores, Helio Duarte
• DFTB
• Thomas Heine, Gotthardt Seifert, Serguei
  Patchkovskii, Helio Duarte,Annick Goursot
• DFT-D, DFTB-D, meta-GGA for damped dispersion
• Petr Jurecka, Pavel Hobza, Jiri Cerny, Yue
  Zhang, Alberto Vela, Thomas Heine, Annick Goursot
• CCM Cyclic Cluster Model
• Florian Janetzko, Andreas Koester
• Kinetic networks
• Rui Zhu, Andre Ribeiro, Stu Kauffman,
• ReaxFF
• Rui Zhu, Adri van Duin, Bill Goddard

17
1200 atoms
800 atoms
10 Opteron 2.4 GHz
400 atoms
18
750 atoms
Nano and Bio converge
19
Development of the ReaxFF Reactive Force Field
for Describing Transition Metal
CatalyzedReactions, with Application to the
Initial Stages of the Catalytic Formation of
Carbon NanotubesKevin D. Nielson, Adri C. T. van
Duin, Jonas Oxgaard, Wei-QiaoDeng, and William A.
Goddard IIIJ. Phys. Chem. A 2005, 109, 493-499

• ReaxFF force field
• Parameters fit to substantial data base of QM
  data, ground state as well as full reactive
  pathways
• Hydrocarbons, nitramines, silicon/silicon oxide,
  aluminum/aluminum oxide, carbon materials, Co,
  Ni, Cu
• Working on Mg, phosphates

20
ReaxFF, continued - J. Phys.Chem A 105(2001)9396

• QM calculations DFT B3LYP
• Based on bond order/bond distance relationship

21
ReaxFF, continued
Corrects for erroneous prediction of previous
ReaxFF that C2 has strong triple bond
22
ReaxFF, continued
23
ReaxFF All carbon compounds
24
ReaxFF metal-molecule interactionsa sample
25
ReaxFF MD for initial stages of nanotube
formation

• NVT MD 1500K, 20x20x20A box, 5C20, 10C4
• without M or with 15 Cu, Ni, or Co

26
ReaxFF MD for initial stages of nanotube
formation
27
ReaxFF MD for initial stages of nanotube
formation
28
Kinetic Monte Carlo a glimpsePHYSICAL REVIEW B
73, 045433 2006First-principles kinetic Monte
Carlo simulations for heterogeneous catalysis
Application to the CO oxidation at
RuO2110Karsten Reuter and Matthias Scheffler

• Sets of coupled reactions 26 in this case
• Rates for elementary processes

29
K. Reuter, C. Stampfl, M. Scheffler, Handbook of
Materials Modeling, Sidney Yip (ed) , Springer,
Berlin (2005) pp 149-234
30
(No Transcript)
31
Steady state surface structures and turnover
frequencies

• Structures and TOF as a function of partial
  pressures of reactants
• Good agreement with experiment
• 5 or 6 of the 26 reactions important
• Sensitivity analysis - The reaction with the
  lowest barrier makes only a minor contribution

32
Does phenomenological kinetics provide an
adequate description of heterogeneous catalytic
reactions?THE JOURNAL OF CHEMICAL PHYSICS 126,
204711 (2007)Burcin Temel, Hakim Meskine,
Karsten Reuter,Matthias Scheffler, Horia Metiu
33
Two examples of stochastic influences on phenotype

• (A) The fingerprints of identical twins. (B) The
  first cloned cat (left) and its genetic mother
  (right). From J.M. Raser, E.K. OShea Science
  309, 2010, 2005

34
Real-time monitoring the single-gene expression
in E. coli under the control of repressed promoter
Rui Zhu, Andre Ribeiro, Stu Kauffman
From J. Yue et al. Science 311, 1600, 2006
35
A model of gene expression
(1)
(4-5)
(2)
RBS
(3)
Protein molecule
Rui Zhu Andre Ribero Stu Kauffman
36
Comparison between simulations and experiments
37
A closer look at transcription

• From S. Greive, P. H. von Hippel Nature Rev 6,
  221, 2005

38
Gene expression
RNA polymerase
Active-site core
39
The two-metal-ion mechanism
Main Issue How does the proton of the 3-OH
group leave?
40
Methods

• I. DFT-level modeling
• Deriving an initial geometry of the active site
  from PDB1I6H
• Optimized by deMon2k at the PBE96-PBE/DZVP level

41
The active-site model
The triad of aspartate residues (1I6H)
The optimized loop structure
42
Methods

• I. DFT-level modeling
• Deriving an initial geometry of the active site
  from PDB1I6H
• Optimized by deMon2k at the PBE96-PBE/DZVP
  level
• Performing a MD simulation in gas phase at 310
  K for 5 ps
• II. ReaxFF-level modeling
• Performing a MD simulation in gas phase at 310
  K for 5 ps

43
MD simulations
44
Methods

• I. DFT-level modeling
• Deriving an initial geometry of the active site
  from PDB1I6H
• Optimized by deMon2k at the PBE96-PBE/DZVP
  level
• Performing a MD simulation in gas phase at 310
  K for 5 ps
• II. ReaxFF-level modeling
• Performing a MD simulation in gas phase at 310
  K for 5 ps
• Building a reactant-side model of 250 atoms
  from PDB2NVZ
• Put in a cell containing 400 water molecules
• Energy minimized and next equilibrated at 310 K
• Cooled down to 10 K
• Driving the reaction coordinate with shifting
  bond restraints

45
Nucleotidyl transfer mechanism 1
Step 1. The polymerase translocates to the next
Addition site the 3-OH group of the just added
nucleotide slides over the active-site loop Step
2. A matched NTP entering the Addition site forms
a reactant-side structure I Step 3. The 3-OH
group performs a nucleophilic attack at the same
time, the proton of the 3-OH group is removed by
Asp485 in a coordinated fashion, forming a
pentacoordinated intermediate II Step 4. The PO
bond cleavage of the intermediate forms a
phosphodiester bond. This leads to a product-side
structure III Go back to step 1 for the next
addition cycle.
46
Nucleotidyl transfer mechanism 2
Deprotonation by bulk water
47
ReaxFF reaction energetic profiles of the two
investigated reaction pathways
48
Whats next?

• More tests with deMon and ReaxFF
• Improved QM/MM in deMon (polarizable FF)
• Learn Kinetic Monte Carlo (multi-scale)
• Calculate multi-scale reaction probabilities
  using deMon and/or ReaxFF energetics and KMC with
  sensitivity analysis.
• The origins of life
• Predictive, atomistic, genetic regulatory networks

49

• Complex systems
• Break them up
• MULTI-SCALE
• Use many techniques
• Make the techniques work together INTEGRATE!

QM
QM
MM

etc.
Q-chem
Stat-mech
dynamics
FE,CFD,