1. Comparative Study of NAMD and GROMACS
Yanbin Wu, Joonho Lee and Yi Wang
Team Project for Phy466
May 11, 2007

2. Outline Motivation Simulation Set-up Procedure Result and analysis
Conclusion

3. Motivation (1) NAMD and GROMACS GROMACS: developed in the Netherlands.
Fast, free.
NAMD:
developed in Urbana, IL.
Parallel, fast for big systems, free.

4. Motivation (2) Compare two packages
Both are widely used MD packages.
Different code implementation in the two packages may cause different results
Generally, one group mainly uses one package
Good chance to compare two packages !!!

5. Simulation Setup (1) Simulation RESULTS Package force field running
parameter
Implementation
Simulation
Package
RESULTS
model
(topological)
force field

6. Simulation Setup (2) Algorithm Initial system Force field Model
Running parameter
Initial system
Size, composition
Coordinate, velocity
Force field
LJ parameter
Model
Charge, bonding, angle parameter
Code implementation
Black box

7. Procedure Find a zero point Compare different water models
NVE
NVT
Compare different water models
TIP3P
SPCE
Compare different temperature control schemes
Langevin
Nose Hoover
Berendsen

8. Simulation system The simplest system Ensembles Water Water+Ions NVE
Solvent of life
Simple & isotropic
Rich in experimental data
Water+Ions
Ensembles
NVE
NVT
Langevin, Nose-hoover, Berendsen

9. Water Model: SPCE SPC/E rigid model (Berendsen et al., 1987)
q(h) = , q(O) =
O-H distance = 1 (Å)
H-O-H angle = °
LJ parameter
A = (kJ/mol)1/6.nm and B = (kJ/mol)1/12.nm

10. Water Model: TIP3P LJ parameter
TIP3P flexible model (Mahoney and Jorgensen, 2000)
q(h) = 0.417, q(O) =
O-H distance = (Å)
H-O-H angle = °
LJ parameter
= , = (Å)

11. Temperature control schemes
Langevin
Introduce a random force and friction coefficient
Nose-hoover
Introduce a thermal reservoir and a friction term in the eq. of motion
Berendsen
Weak coupling first-order kinetics to an external heath bath with a given temperature

12. Results and Analysis (1)
Zero point
NVE
0.25 ns NVT to bring temperature up to 300K.
1 ns NVE.
Important: start with the same velocity and coordinate in both packages.
NVT
1 ns NVT using Langevin dynamics temperature control, damping coefficient 5/ps.

13. NVE Zero Point 5.6% difference GROMACS NAMD Water model Tip3p
Diffusion coefficient (10-5cm2/s)
(+/ )
5.278 (+/-0.012)
5.6%
difference

14. NVT Zero Point 7.3% difference GROMACS NAMD Water model Tip3p
Temperature control scheme
Langevin (5)
Diffusion coefficient (10-5 cm2/s)
(+/ )
2.769 (+/-0.002)
7.3%
difference

15. Results and Analysis (2)
Different water models
1 ns NVT using Langevin dynamics with  = 5/ps.
Two different water models: SPCE and TIP3P.
Most commonly used water models.
*SPCE and TIP3P water models exhibit similar dynamic properties in GROMACS. 
SPCE
TIP3P 5
(+/ )
(+/ )
0.032% difference

16. Results and Analysis (3)
Damping in Langevin Dynamics
1 ns NVT via Langevin dynamics with  = 1, 5, 10 /ps.
*Damping affects the diffusion of water dramatically.
*=5/ps best reproduces the experimental result. 
GROMACS (TIP3P)
NAMD (TIP3P) 1
(+/ )
4.259 (+/-0.025) 5
(+/ )
2.769 (+/-0.002) 10
(+/ )
1.958 (+/-0.003)

17. Results and Analysis (4)
Different temperature control schemes
1 ns NVT using Langevin dynamics with  = 5/ps
1 ns NVT using Nose-hoover thermostat with =0.1 ps
1 ns NVT using Berendsen thermostat with =0.1 ps
*Different temperature control schemes can achieve similar results with well-chosen parameters.
GROMACS(SPCE)
Diffusion coefficient (10-5 cm2/s)
Langevin
(+/ )
Berendsen
(+/ )
Nose-Hoover
(+/ )

18. Results and Analysis (5)
Water in the water+ion system
Package
Temperature
control scheme
water model /
force field
Diffusion coefficient (10-5 (cm2/s))
NAMD
Langevin (5)
tip3p (namd) /
CHARMM
(+/ )
GROMACS
Berendsen
spce(gromacs)/
(+/ )
spce(reference)
/ reference
(+/ )
Nose-hoover
(+/ )
(+/ )

19. Results and Analysis (5)
Na+ in the water+ion system
Package
Temperature
control scheme
water model /
force field
Diffusion coefficient (10-5 (cm2/s))
NAMD
Langevin (5)
tip3p (namd) /
CHARMM
(+/ )
GROMACS
Berendsen
spce(gromacs)/
(+/ )
spce(reference)
/ reference
(+/ )
Nose-hoover
(+/ )
(+/ )

20. Results and Analysis (5)
Cl- in the water+ion system
Package
Temperature
control scheme
water model /
force field
Diffusion coefficient (10-5 (cm2/s))
NAMD
Langevin (5)
tip3p (namd) /
CHARMM
(+/ )
GROMACS
Berendsen
spce(gromacs)/
(+/ )
spce(reference)
/ reference
(+/ )
Nose-hoover
(+/ )
(+/ )

21. Results and Analysis (5)
Radial distribution of oxygen - oxygen

22. Results and Analysis (5)
Radial distribution of oxygen – Cl-

23. Results and Analysis (5)
Radial distribution of oxygen – Na+

24. Conclusion
The two packages GROMACS and NAMD produce similar results (within tolerance) using the same set of parameters.
Damping coefficient affects the dynamics significantly and has to be chosen with caution.
Different temperature control schemes may generate similar dynamic properties.
Two different water models, SPCE and TIP3P were compared and only minor difference was observed regarding the diffusion of water.

25. Discussion Energy conservation in NVE simulations
Neighbor list update frequency
Switch or shift function is required, instead of cutoff
PME
Damping coefficient in NVT simulations
Balance temperature fluctuation and disturbance to the motion of the system.
Different temperature control schemes

26. Zero point (water+ion, NVT)
GROMACS
NAMD
Forcefield
CHARMM
watermodel(bonding)
tip3p(namd)
Temperature control scheme
Langevin (5)
Diffusion coefficient (cm2/s)
Water
(+/ )
(+/ )
Na+
(+/ )
(+/ )
Cl-
(+/ )
(+/ )

27. Water Model (3) Package NAMD 2.4895e-3 2.4248e-6 3.1540e-9 1.2921e-17
TIP3P
(NAMD)
O / H
2.4895e-3
2.4248e-6
3.1540e-9
1.2921e-17
GROMACS
(GROMACS)
2.4889e-2
2.4352e-6
SPCE
2.6171e-3
2.6331e-6
(reference)
2.6341e-3
2.6679e-6

28. NVE Zero Point GROMACS NAMD Forcefield CHARMM watermodel(bonding)
tip3p(namd)
Diffusion coefficient (cm2/s)
(+/ )
5.278 (+/-0.012)