1. implicit vs. explicit solvation
water is only included implicitly as an effective (averaged) interaction with no atomic detail
an explicit representation of water is included at the molecular level
What does solvent do?

2. implicit vs. explicit solvation
water is only included implicitly as an effective (averaged) interaction with no atomic detail
an explicit representation of water is included at the molecular level
What does solvent do?
viscous damping force
dielectric screening / polarization
hydrophobic effect
structure

3. implicit or no solvent

4. U = q e r e = 1 gas phase e = 80 liquid phase ? e = 4pe0e
What does solvent do?
viscous damping force (not important)
dielectric screening  U = q i j e r ij
e = 1 gas phase
e = 80 liquid phase ?
dielectric constant,
e = 4pe0e

5. This is BULK solvent screening. At short range, no screening…
What does solvent do?
viscous damping force (not important)
dielectric screening  U = q i j e r ij
e = 1 gas phase
e = 80 liquid phase ?
This is BULK solvent screening. At short range, no screening… 80 e
vs. 1
distance

6. U = q e r e = rij or 4rij e What does solvent do?
viscous damping force (not important)
dielectric screening  U = q i j e r ij
Simple approximation:
Distance dependent dielectric
e = rij or 4rij 80 e 1
distance

7. U = q e r = rij or 4rij e What does solvent do?
viscous damping force (not important)
dielectric screening  U = q i j e r ij
Simple approximation:
Distance dependent dielectric
= rij or 4rij
Better? Sigmoidal dielectric 80 e 1

8. What does solvent do?
viscous damping force (not important)
dielectric screening 
hydrophobic effect
What does solvent do?
viscous damping force (not important)
dielectric screening 

9. surface area approaches
Observation: Gsolvation for the saturated hydrocarbons in water is linearly related to the solvent accessible surface area
…the line passing through the nonpolar Ala, Val, Leu and Phe has a slope of 22 cal/Å2
First parameterization:
-- D. Eisenberg & A. D. McLachlan, “Solvation energy in protein folding and binding.” Nature 319, (1986).
The Scheraga hydration shell model:
-- T. Ooi, M. Oobatake, G. Nemethy & H. A. Scheraga, “Accessible surface area as a measure of the thermodynamic parameters of hydration of peptides.” Proc. Natl. Acad. Sci. 84, (1987).
-- Y. K. Kang, K. D. Gibson, G. Nemethy & H. A. Scheraga, “Free energies of hydration of solute molecules. 4. Revised treatment of the hydration shell model.” J. Phys. Chem. 92, (1988).
-- R. L. Williams, J. Vila, G. Perrot & H. A. Scheraga, “Empirical solvation models in the context of conformational energy searches: Application to bovine pancreatic trypsin inhibitor.” Proteins 14, (1992).
Use in molecular dynamics simulations:
-- C. A. Schiffer, J. W. Caldwell, P. A. Kollman & R. M. Stroud, “Protein structure prediction with a combined solvation free energy-molecular mechanics force field.” Mol. Sim. 10, (1993).
Creighton Proteins (FM Richards, Ann Rev Biophys Bioeng 6, (1977))

10. surface area approaches
Observation: Gsolvation for the saturated hydrocarbons in water is linearly related to the solvent accessible surface area
Problems:
sensitive to si’s, parameterization, surface area and change in conformation
in dynamics you need derivatives of SASA
what about polarization effects?
exposed solvent accessible surface area (SASA)
DGresidue =  DsiAi
atoms,i
free energy of interaction of a solute with water
atomic solvation parameters based on free energies of transfer
First parameterization:
-- D. Eisenberg & A. D. McLachlan, “Solvation energy in protein folding and binding.” Nature 319, (1986).
The Scheraga hydration shell model:
-- T. Ooi, M. Oobatake, G. Nemethy & H. A. Scheraga, “Accessible surface area as a measure of the thermodynamic parameters of hydration of peptides.” Proc. Natl. Acad. Sci. 84, (1987).
-- Y. K. Kang, K. D. Gibson, G. Nemethy & H. A. Scheraga, “Free energies of hydration of solute molecules. 4. Revised treatment of the hydration shell model.” J. Phys. Chem. 92, (1988).
-- R. L. Williams, J. Vila, G. Perrot & H. A. Scheraga, “Empirical solvation models in the context of conformational energy searches: Application to bovine pancreatic trypsin inhibitor.” Proteins 14, (1992).
Use in molecular dynamics simulations:
-- C. A. Schiffer, J. W. Caldwell, P. A. Kollman & R. M. Stroud, “Protein structure prediction with a combined solvation free energy-molecular mechanics force field.” Mol. Sim. 10, (1993).

11. what is the effective solvent polarization? (solve Poisson equation)
Born: isolated point charge (q) in a spherical cavity of radius r immersed in a dielectric continuum with dielectric constant e ? r
einside
eoutside q

12. what is the effective solvent polarization? (solve Poisson equation)
Born: isolated point charge (q) in a spherical cavity of radius r immersed in a dielectric continuum with dielectric constant e

13. what is the effective solvent polarization? (solve Poisson equation)
Born: isolated point charge (q) in a spherical cavity of radius r immersed in a dielectric continuum with dielectric constant e
Onsager: neutral system with dipole m

14. GBSA: generalized Born surface area approach{
SASA term
Gsolvation = Gcavity + Gvdw + Gpolarization ? m q qi qj

15. GBSA: generalized Born surface area approach{
SASA term
Gsolvation = Gcavity + Gvdw + Gpolarization
e: dielectric constant, qi’s: charges, rij: distance between atom pairs.
ai: Born radii; calculated numerically for each charged atom in solute; values change as calculation proceeds.
k: modification to incorporate salt effects at low salt via a Debye-Huckel term
- ( )qiqj/fGB
e-kfGB e 1 2
fGB = (r2ij + a2ije-Dij)1/2
aij = (aiaj)1/2
Dij = r2ij/ 2a2ij

16. GBSA: generalized Born surface area approach{
SASA term
Gsolvation = Gcavity + Gvdw + Gpolarization
e: dielectric constant, qi’s: charges, rij: distance between atom pairs.
ai: Born radii; calculated numerically for each charged atom in solute; values change as calculation proceeds.
k: modification to incorporate salt effects at low salt via a Debye-Huckel term
“this expression gives the Born equation for superimposed charges when rij = 0, the Onsager reaction field within 10% for a dipole in a spherical cavity when rij < 0.1 aij, and the Born plus Coulomb dielectric polarization energy within 1% for two charged spheres when rij > 2.5 aij.” (Still et al., 1990).
- ( )qiqj/fGB
e-kfGB e 1 2
fGB = (r2ij + a2ije-Dij)1/2
aij = (aiaj)1/2
Dij = r2ij/ 2a2ij
One of the most widely used and cited implementations of the generalized Born methodology is within the Macromodel program using the implementation by Still’s group. In molecular dynamics, the Born radii are assumed fixed over certain intervals (simplifying the derivatives for speed). This effectively reduces the molecular dynamics simulations to running with an effective dielectric constant (unless the radii are allowed to change at every step of the simulation which tremendously increases the cost).
-- W. C. Still, A. Tempczyk, R. C. Hawley & T. Hendrickson, “Semianalytic treatment of solvation for molecular mechanics and dynamics.” J. Amer. Chem. Soc. 112, (1990) & supplementary material
This isn’t actually the original citation as the method goes back to earlier times…
-- G. J. Hoijtink, E. De Boer, P. H. van der Meij & W. P. Weijland, “Reduction potentials of various aromatic hydrocarbons and their univalent anions.” Recueil des travaux chimiques des Pays-Bas 75, (1956).
Recently the method has seen a resurgence thanks to more elegant methods to estimate the Born radii using a pairwise descreening approximation (PDA) from a sum over atom pairs which simplifies the derivatives and allows evolution of the Born radii as the simulation proceeds. The PDA is discussed by Hawkins & Cramer:
-- G. D. Hawkins, C. J. Cramer & D. G. Truhlar, “Pairwise solute descreening of solute charges from a dielectric medium”, Chem. Phys. Lett. 246, (1995).
-- G. D. Hawkins, C. J. Cramer & D. G. Truhlar, “Parameterized models of aqueous free energies of solvation based on pairwise descreening of solute atomic charges from a dielectric medium”, J. Phys. Chem. 100, (1996).
With appropriate parameterization of the individual Born radii (and atomic solvation parameters), stable simulations of proteins and nucleic acids have been observed.
-- V. Tsui & D. A. Case, “Molecular dynamics simulations of nucleic acids with a generalized Born solvation model”, J. Amer. Chem. Soc. 122, (2000).
-- D. J. Williams & K. B. Hall, “Unrestrained stochastic dynamics simulations of the UUCG tetraloop using an implicit solvation model”, Biophys. J. 76, (1999).
-- B. N. Dominy & C. L. Brooks, III, “Development of a generalized Born model parameterization for proteins and nucleic acids”, J. Phys. Chem. B. 103, (1999).

17. GBSA: generalized Born surface area approach
Excellent agreement to FEP for neutral small molecules with only ~2-4x overhead compared to gas phase dynamics
Implemented in MacroModel, AMSOL, AMBER and CHARMM. [Recent resurgence in interest! Stable MD simulation / folding !!!]
Can give stable trajectories of proteins and nucleic acids; ~2-5x cost of in-vacuo simulations.

18. Poisson-Boltzmann electrostatics
treats the solute as a low (fixed) dielectric medium containing atomic charges surrounded by a molecular surface immersed in a dielectric continuum
electrostatic component from solution of PB equations; most often done numerically by finite difference
hydrophobic/cavity term from surface area
gives reasonable free energies of solvation
can include salt effects through solution of non-linear PB
Fixed, low dielectric in interior. Charges located at atomic positions. Surface (or dielectric discontinuity) is defined by SASA or vdw surface or …
For review see:
B. Honig & A. Nicholls, “Classical electrostatics in biology and chemistry”, Science 268, (1995).
M. E. Davis & J. A. McCammon, “Electrostatics in biomolecular structure and dynamics”, Chem. Rev. 90, (1990).
K. A. Sharp & B. Honig, “Electrostatic interactions in macromolecules: theory and applications.” Ann. Rev. Biophys. Biophys. Chem. 19, (1990).
J. D. Madura, J. M. Briggs, R. C. Wade, M. E. Davis, B. A. Luty, A Ilin, J. Antosiewicz, M. K. Gilson, B. Bagheri, L. R. Scott & J. A. McCammon, “Electrostatics and diffusion of molecules in solution: simulations with the University of Houston Brownian Dynamics program.” Comp. Phys. Comm. 91, (1995).
Dielectric continuum

19. Poisson-Boltzmann electrostatics
treats the solute as a low (fixed) dielectric medium containing atomic charges surrounded by a molecular surface immersed in a dielectric continuum
electrostatic component from solution of PB equations; most often done numerically by finite difference
hydrophobic/cavity term from surface area
gives reasonable free energies of solvation
can include salt effects through solution of non-linear PB
Issues:
no microscopic detail
strong dependence on radii
time consuming in MD
sensitivity to grid
sensitivity to interior/exterior dielectric
Programs: DELPHI (Honig/Sharp), MEAD (Bashford), UHBD (McCammon), GRASP (Nicholls), APBS, ...
For review see:
B. Honig & A. Nicholls, “Classical electrostatics in biology and chemistry”, Science 268, (1995).
M. E. Davis & J. A. McCammon, “Electrostatics in biomolecular structure and dynamics”, Chem. Rev. 90, (1990).
K. A. Sharp & B. Honig, “Electrostatic interactions in macromolecules: theory and applications.” Ann. Rev. Biophys. Biophys. Chem. 19, (1990).
J. D. Madura, J. M. Briggs, R. C. Wade, M. E. Davis, B. A. Luty, A Ilin, J. Antosiewicz, M. K. Gilson, B. Bagheri, L. R. Scott & J. A. McCammon, “Electrostatics and diffusion of molecules in solution: simulations with the University of Houston Brownian Dynamics program.” Comp. Phys. Comm. 91, (1995).

20. implicit METHODS: modified dielectric functions “surface area” methods
“continuum” or Poisson-Boltzmann
generalized Born methods
[reaction field, integral equation methods]
BENEFITS:
Averaging is implicit; gives free energy of solvation
Conformational sampling is “faster”

21. implicit
CURSE:
cannot represent “specific” (direct) structural water interaction
the computational cost is still large
force field balance / parameterization
methods still under development (i.e. in flux)
METHODS:
modified dielectric functions
“surface area” methods
“continuum” or Poisson-Boltzmann
generalized Born methods
[reaction field, integral equation methods]
BENEFITS:
Averaging is implicit; gives free energy of solvation
Conformational sampling is “faster”

22. explicit solvation ISSUES: water model sampling
viscous damping
dielectric screening
hydrophobicity, …, ?
ISSUES:
water model
sampling
boundary conditions (periodic or non-)
long range forces, cutoff methods/effects
METHODS:
[Langevin dipole]
TIP3P and other models (SHAKE)

23. explicit solvation what water model to use?
what representation or boundary conditions?
how to make the calculation tractable?
how much water to add?

24. Explicit water models flexible water models: few are in general use…
(see Levitt et al., 1995; Ferguson, 1995; Mizan et al., 1994, etc.)
rigid water models:
require SHAKE/RATTLE to constrain bonds and angles
TIP3P: method of choice with Cornell et al. force field
[well balanced with 6-31 G* charges since prepolarized through larger fixed dipole]
TIP4P/TIP5P: supported within LEaP (and PLEP)
SPC/E: works as well as TIP3P (or better), not supported by default in AMBER.
TIP3P, TIP4P: Jorgensen et al., 1983
SPC/E: Berendsen et al., 1987
TIP: r(OH) = Å, >HOH = º
SPC/E: r(OH) = 1.0 Å, >HOH = º
TIP3P: W. L. Jorgensen, J. Chandrasekhar & J. D. Madura, “Comparison of simple potential functions for simulating liquid water.” J. Chem. Phys. 79, (1983)
SPC/E: H. J. C. Berendsen, J. R. Grigera & T. P. Straatsma, “The missing term in effective pair potentials.” J. Phys. Chem. 91, (1987).

25. rigid water models: how do they perform?
TIP3P
SPC/E
experiment
r (g/cm3)
[density]
0.982 (*)
0.99 cut
0.97 PME
1.0 cut
0.99 PME
0.997
Ei (kcal/mol)
[interaction energy]
-9.86 (*)
-9.7 cut
-9.5 PME
-11.3 cut
-11.1 PME
-9.92
D (x10-5 cm2/s)
[diffusion]
5.3 cut
5.1 PME
2.5 cut
2.7 PME
2.3 at 25º
“reasonable” density, interaction energies and 1st peak radial distribution function occupancies
3 site models underestimate compressibility
TIP3P has too little structure beyond the 1st peak; less tetrahedrality than TIP4P and tends to diffuse too rapidly
TIP4P has excellent density and proper oxygen structure however hydrogens are somewhat misplaced
In simulations of nucleic acids, TIP3P and SPC/E lead to roughly equivalent hydration patterns

26. explicit solvation what water model to use?
what representation or boundary conditions?
how to make the calculation tractable?
how much water to add?

27. explicit water boundary conditions
finite vs. infinite
surrounding? (vacuum, continuum, ...)
size and shape (sphere, orthorhombic, truncated octahedral)

28. explicit water boundary conditions
“blob” or cap
truly periodic or Ewald

29. vacuum boundary conditions
spherical shell of atoms around the site of interest
Problems:
- surface tension leads to high pressure
- reduced fluctuations
- vacuum interface, waters can drift...
“blob” or
cap

30. vacuum boundary conditions
spherical shell of atoms around the site of interest
Problems:
- surface tension leads to high pressure
- reduced fluctuations
- vacuum interface, waters can drift...
“blob” or
cap
alternatives: stochastic boundary …
Langevin forces on “near” surface waters
surface waters held fixed at “bulk” density

31. vacuum boundary conditions
spherical shell of atoms around the site of interest
Problems:
- surface tension leads to high pressure
- reduced fluctuations
- vacuum interface, waters can drift...
“blob” or
cap
alternatives: solvent boundary potentials
how to develop??? +

32. vacuum boundary conditions
spherical shell of atoms around the site of interest
Problems:
- surface tension leads to high pressure
- reduced fluctuations
- vacuum interface, waters can drift...
“blob” or
cap
alternatives: reaction field or dielectric continuum
What happens to waters that leave?
(explicit to implicit conversion?)

33. explicit water boundary conditions
truly periodic or Ewald

34. explicit solvation what water model to use?
what representation or boundary conditions?
how to make the calculation tractable?
how much water to add?

35. nonbonded
interactions
Can we speed up the calculations by limiting the number of pair interactions?

36. CUTOFFS x x minimum image spherical reorientational motion slowed!
Roberts & Schnitker, 1995

37. nonbonded
interactions

39. nonbonded
interactions

40. 10.0 kcal at 10 Å!

41. nonbonded
interactions

42. Can we avoid artifacts due to the cutoff of pair interactions?
Avoid minimum image
Don’t split dipoles: charge group neutrality (AMBER)
Use two cutoffs: CUT2ND>CUT
Do not ignore long ranged electrostatic interactions
Ewald (Ewald, 1921), particle mesh Ewald (Essman et al., 1995; Darden & Sagui, 2000), particle-particle particle-mesh Ewald (Luty & van Gunsteren, 1996)
Fast multipole method (Greengard & Rokhlin, 1989; Board et al., 1992; Lambert et al., 1996)
Cell multipole (Ding et al., 1992)
outer interactions are updated less frequently x
Truncated “group-based” cutoffs have traditionally been applied within AMBER. The idea was that artifacts due to the truncation could be minimized by avoiding the splitting of dipoles. This assumption turned out to be in error and in fact, very deleterious artifacts can be seen with the application of group-based truncation. Better (as discussed in the next slide) is to avoid the energy and force discontinuities through the application of a switched or shifted potential/force. Even better is to explicitly include the long-ranged electrostatic interactions. Recently an excellent review by Darden & Sagui has appeared discussing long-range electrostatic methods; this reference contains a discussion of the work cited above.
C. Sagui, T. A. Darden, “Molecular dynamics simulations of bio-molecules: Long-range electrostatic effects.” Ann. Rev. Biophys. Biomol. Struct. 28, (1999).
T. E. Cheatham, III & B. R. Brooks, “Recent advances in molecular dynamics simulation towards the realistic representation of biomolecules in solution.” Theor. Chem. Acc. 99, (1998).

43. improving cutoffs? switch shift
shift/switch the energy or force discontinuity

44. shift/switch the energy or force discontinuity
improving cutoffs?
shift/switch the energy or force discontinuity
Better to use longer cutoff than worry about details of switch/shift
Best cutoff method is atom-based force shift
Switch is not-advised for electrostatics
Truncation may lead to “heating” due to dipole reorientation near the cutoff
shift
switch

45. …but what about artifacts from the cutoff?
Examples of the deleterious effects of the cutoff
DNA falls apart (Cheatham et al., 1995)
electrostatic potential (Pettitt & Smith, 1991)
anomalous water transport (Feller et al., 1996)
alpha helical peptides (Schreiber & Steinhauser, 1992)
attractive ion PMF (Bader & Chandler, 1992)

46. d[CCAACGTTGG]2 120 ps, 12 Å cutoff 50 ps, 16 Å cutoff
overall extension of duplex, middle base pairs fray
terminal base pairs fray, duplex bends
120 ps, 12 Å cutoff ps, 16 Å cutoff
(group based truncation)

47. Coulombic potential atom switch group switch Ewald
Smith & Pettitt J. Chem. Phys. 95, (1991)
Coulombic potential
atom
switch
group switch
Ewald

48. Ewald (solid) 12 Å force shift 10-12 Å force switch
S. E. Feller, R. W. Pastor, A. Rojnuckarin, S. Bogusz, B. R. Brooks, “Effect of electrostatic force truncation on interfacial and transport properties of water.” J. Phys. Chem. 100, (1996).
density weighted water polarization profile
Electrostatic potential profile across vapor/water/vapor interface
Ewald
Ewald (solid)
12 Å shift
12 Å force shift
10-12 Å force switch

49. Biochemistry 31,

50. W(R): PMF of two ions Fe2+-Fe3+ CUTOFF Fe2.5+-Fe2.5+ EWALD +
J. Phys. Chem. 96, (1992)
W(R): PMF of two ions
Fe2+-Fe CUTOFF
Fe2.5+-Fe2.5+ EWALD
Ewald
cutoff
non-integral charge to avoid dipole correction; leads to same structure and energetics
not cutoff spline dependent!
Artifact is > 8 kBT!!!
in simulation with cutoffs, two like charged ions moved closer together… +
dielectric continuum

51. Periodic boundary conditions: Ewald electrostatics
truly periodic or Ewald

52. å qi qj electrostatics... 1 2 __ + UQcorr | ri - rj | UQ = U rijN
(assume e=1)
rij < cutoff

53. is conditionally convergent
electrostatics... U
rij
boundary conditions? ...extend to truly periodic
n: sum over lattice vectors
n = nxX + nyY + nzZ
implies n = 0 & i = j omitted 1 2 __ å qi qj +
UQcorr
| ri - rj |
UQ =
i,j=1 N
(assume e=1)
rij < cutoff 1 2 __ å qi qj
| ri - rj + n |
UQ =
i,j=1 N n , å n , ,
is conditionally convergent

54. Udirect: Ewald UQ = Udirect +Ureciprocal +U self +Usurface charges
“j” periodic image
“j”
“i”
Udirect:

55. Udirect: Ewald UQ = Udirect +Ureciprocal +U self +Usurface r(r) = b3ex
erfc(x) = e-t2dt 2
p1/2
UQ = Udirect +Ureciprocal +U self +Usurface
charges
“j” periodic image
“j”
“i”
Udirect:
-b2r2
p3/2
r(r) = b3e
r: position relative
to center
b: width
screening “counterion”
charge density, r(r)

56. Udirect: Uidirect= qj Ewaldx
erfc(x) = e-t2dt 2
p1/2
UQ = Udirect +Ureciprocal +U self +Usurface
charges
“j” periodic image
“j”
“i”
Udirect:
-b2r2
p3/2
r(r) = b3e
r: position relative
to center
b: width
screening “counterion”
charge density, r(r)
Uidirect= qj
erfc(b|ri-rj|)
|ri-rj|
...extend to periodic system, add “n” and sum over all charges:

57. Simplify? adjust b so only terms within the cutoff are significant...
Ewald x
erfc(x) = e-t2dt 2
p1/2
UQ = Udirect +Ureciprocal +U self +Usurface
charges
“j” periodic image
“j”
“i”
Udirect:
-b2r2
p3/2
r(r) = b3e
r: position relative
to center
b: width
screening “counterion”
charge density, r(r)
Uidirect= qj
erfc(b|ri-rj|)
|ri-rj|
...extend to periodic system, add “n” and sum over all charges:
Udirect= qiqj
erfc(b|ri-rj+n|)
|ri-rj+n| 1 2 __ å
i,j=1 N n ,
Simplify? adjust b so only terms within the cutoff are significant...

58. Ureciprocal: Ewald: “reciprocal sum”
...solution to Poisson’s equation with periodic boundary conditions and source term r(r)
[expand in Fourier series, solve term by term...]

59. å å | S(m) | e 2pim.|ri| Ureciprocal: e -p2m2/b2 = c.F(Q) S(m) =
Ewald: “reciprocal sum”
Ureciprocal:
...solution to Poisson’s equation with periodic boundary conditions and source term r(r)
[expand in Fourier series, solve term by term...]
Ur = 1
2pV __ å
m=0
e -p2m2/b2 m2
| S(m) | 2
V: volume of unit cell
m: mxX*+myY*+MzZ* å
i=1 N qi
e 2pim.|ri|
S(m) =
= c.F(Q) ?
The key to the speed is the FFT on Q, the cardinal spline interpolated charge grid...

60. particle mesh Ewald: other terms...
Uself = S qi2 -b
p1/2
self term:
“surface” term: assume spherical boundary
(dipole, surroundings, “shape”)
“tin foil boundary”: e = ¥ implies Usurface = 0
PME extensions...
r-6 “Ewald” (particle mesh)
PME into gibbs (free energy simulation)
further parallelization and single PE optimization, beyond SPMD?
Full conversion to MPI and distribution of source with AMBER 4.x
PME by default in AMBER6
PME with dipoles, quadrupoles (i.e. polarization)
Usurface= S qi.ri2 2p 3v

62. Consequences of imposing true periodicity
dipole reorientation
vs.
induced correlations
vs. + -

63. Consequences of imposing true periodicity
dipole reorientation
vs.
induced correlations - +
vs. - + - + - +
surface effects?
tin-foil vs. vacuum boundary conditions
non-ideal (effectively increased) molality
anisotropic interaction energy
uniform neutralization background, DG
force field artifacts?

64. Consequences of imposing true periodicity
Smith & Pettitt: true periodicity is not a problem in solvents with a reasonably high dielectric!
dipole rotation
protein rotation (effect < kT)
Cheatham & Kollman, Norberto de Souza & Ornstein
no apparent dependence on box size on DNA
5 Å, 10 Å or 15 Å of water surrounding the DNA
little salt effects
….but:

65. Consequences of imposing true periodicity
20 Å box
30 Å box
40 Å box
artificial stabilization of a-helix
poly-alanine octapeptide,
2 ns simulations
Hünenberger & McCammon
J. Phys. Chem. B. 104, (2000)
Continuum calculations show artifacts,
Confirmed in MD simulations

66. Cutoff or Ewald?
If using a cutoff, use a “good” cutoff method such as atom-based force shifting.
Be aware of potential artifacts (water transport properties, …)
If using Ewald (i.e. the correct physics) make sure there is enough surrounding solvent or sufficient dielectric to screen image artifacts.
Note: for many systems (such as systems that are not highly charged) these artifacts may not be large!