Skip to main content

gbmv (c47b2)

Generalized Born using Molecular Volume (GBMV)
Solvation Energy and Forces Module
- and -
Surface Area

Questions and comments regarding GBMV should be directed to

Michael S. Lee or Michael Feig c/o
Charles L. Brooks, III (

* Description | Description of GBMV and related commands
* Syntax | Syntax of the GBMV Commands
* Function | Purpose of each of the commands
* Examples | Usage examples of the GBMV module


The GBMV module is a Generalized Born method for
mimicking the Poisson-Boltzmann (PB) electrostatic solvation energy. The PB
method for obtaining solvation energies is considered a benchmark for implicit
solvation calculations. However, the PB method is slow and the derivatives,
i.e. forces, are ill-defined unless one changes the definition of the
molecular volume.
The Generalized Born equation, as prescribed by Still, et. al. allows
one to compute solvation energies very similar to the PB equations.
As it is an analytical expression, forces are available as well:

q q
N N i j
G = -C (1-1/eps){1/2 sum sum ------------------------------------ }
pol el i=1 j=1 [r^2 + alpha *alpha exp(-D )]^(0.5)
ij i j ij

D = r^2 / (K_s * alpha * alpha )
ij ij i j

where K_s = 4 for Still's original equation, or 8 for modified equation.
The only problem is that one needs to calculate the alpha's, a.k.a.
Born radii for each atom, accurately. There are various methods available,
such as the GBORN, ACE, and GBSW modules in CHARMM.
The GBMV method obtains the Born radii very accurately,
i.e, w/ greater than 0.99 correlation. It is available as three approaches:

1) grid-based (Most accurate)
2) analytical method I (Least accurate, fastest)
2) analytical method II (preferred for dynamics)

The analytical method has derivatives and thus can be used in molecular
dynamics simulations. The grid-based method has no derivatives, however,
it is the most accurate and still can be used in energy ranking and
Monte-Carlo methods.

When should you use GBMV?

Because the analytical and grid-based methods are quite accurate, the
parameters change very little when optimized for a particular force-field.
Hence, forcefields besides those of CHARMM can be used with GBMV without
refitting of parameters. The GBMV method II approximates the molecular
surface directly. The agreement with respect to electrostatic solvation
energies from standard Poisson theory is very good (<1% relative error).
The higher accuracy comes at a price, however, and GBMV is slower than
other GB methods in CHARMM.

Papers related to GBMV method:

(1) M. S. Lee, F. R. Salsbury, Jr., and C. L. Brooks III.
J. Chem. Phys.(2002),116, 10606
(2) M. S. Lee, M. Feig, F. R. Salsbury, Jr., and C. L. Brooks III.
J. Comp. Chem. (2003), 24, 1348
(3) M. Feig, A. Onufriev, M. S. Lee, W. Im, D. A. Case, C. L. Brooks III,
J. Comp. Chem. (2004), 25, 265-284
(4) S. Tanizaki, M. Feig,
J. Chem. Phys. (2005), 122, 124706
(5) J. Chocholousova, M. Feig,
J. Comp. Chem. (2006) 27, 719-729
(6) A. Panahi, M. Feig,
J. Chem. Theory Comput. (2013), 9, 1709-1719
(7) B. Dutagaci, M. Sayadi and M. Feig,
J. Comp. Chem. (2017) 38, 1308-1320

Surface Area

A solvent accessible surface area (SASA) calculation is implemented
within the GB module. There is essentially no additional cost compared to the
GB calculation itself and about 5 times faster than an exact, analytical
calculation (e.g. from the ASP module). It is accurate to within 1% of the
exact surface and is much more accurate than the SASA module within CHARMM.

Syntax of Generalized Born Molecular Volume (GBMV) Solvation commands

[SYNTAX: GBMV commands]

[method I: faster but less accurate]

GBMV { P1 <real> P2 <real> LAMBda1 <real> DN <real> SHIFT <real>
WATR <real> BETA <real> EPSILON <real>
SA <real> SB <real> SCUT <real>}
CORR <int>
{ ESHIFT <real> SHIFT <real> TT <real> (CORR = 0) }
{ SHIFT <real> SLOPE <real> (CORR = 1) }

[method II: slower but more accurate (recommended)]

BETA <real> EPSILON <real> DN <real> WATR <real>
LAMBDA1 <real> TOL <real> BUFR <real> MEM <int> CUTA <int>
HSX1 <real> HSX2 <real> ONX <real> OFFX <real>
ALFRQ <int> EMP <real>
P1 <real> P2 <real> P3 <real> P4 <real> P6 <real>
SA <real> SB <real> SCUT <real>
KAPPA <real>
WTYP <int> NPHI <int>
CORR <int>
{ ESHIFT <real> SHIFT <real> TT <real> (CORR = 0,2) }
{ SHIFT <real> SLOPE <real> (CORR = 1,2) }
{ A1 <real> A2 <real> A3 <real> (CORR = 2) }
GCUT <int>
RADG <int> <real ...>
FAST 1|0 SGBFRQ <int> SXD <real>

[HDGB method: heterogeneous dielectric / membrane model]

GBMV { { GBMV II options }
CORR <int> (CORR = 3, 4, 5)
A1 <real> A3 <real> A4 <real> A5 <real>
UNEPS <int>
ZS <real> ZM <real> ZT <real> ST0 <real>

[ grid-based method ]

EPSILON <int> DN <real> WATR <real>
P6 <real>
KAPPA <real>
WTYP <int> NPHI <int>
CORR <int>
{ SHIFT <real> SLOPE <real> (CORR = 1) }
{ ESHIFT <real> SHIFT <real> (CORR = 0) } }

[ free-up memory and/or start over]



Parameters of the Generalized Born using Molecular Volume Model
common to all methods:

WTYP Angular integration grid type:

0 - Dodecahedron
1 - Spherical polar
2 - Lebedev (DEFAULT)
3 - Alternating octahedron/cube

NPHI Used when WTYP equals 1 or 2. When WTYP=1, it corresponds to number
of phi angles. When WTYP=2, it corresponds to size of
Lebedev grid, which can only have values of 6,26 (Default), and 38
at the present time.

CUTA Extent of radial integration points in Angstroms. (Default 20)

GCUT radial spacing of integration grid

1 - default spacing:
0.1 0.2 0.3 0.4 0.5 0.75 1.0 1.25
1.5 1.75 2.0 2.5 3.0 3.5 4.0 5.0
6.0 7.0 8.0 10.0 12.0 16.0 20.0

2 - finer spacing for small radii
0.1 0.2 0.3 0.4 0.5 0.6 0.8 1.0
1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6
2.8 3.0 3.2 3.7 4.1 5.1 6.1 7.1
8.1 10.1 12.1 16.1 20.1

3 - custom grid, specify with RADG

RADG custom grid spacing, first argument is number of intervals
following arguments are interval limits

CORR Coloumb field correction method:

0 for R^5 method. use: SHIFT/ESHIFT/TT

alpha(i) = - 1/( r4 - TT * r5 + ESHIFT ) + SHIFT

1 for R^7 method (default) use: SHIFT/SLOPE

alpha(i) = SLOPE/( (1-1/sqrt(2)) * r4 + r7) + SHIFT

2 for R^5/R^7 use: A1/A2/A3/SHIFT/SLOPE/ESHIFT

alpha(i) = SLOPE/( A1 * r4 + A2 * r5 + A3 * r7 + ESHIFT) + SHIFT

this mode is intended for the calculation of Born radii
in different dielectric environments

3 for R^5/R^7 use: A1/A3/A4/A5/SLOPE

alpha(i) = SLOPE/( A1 * r4 + A3(i) * r7) + A4 + A5/(eps(i) + 1)
A3(i) = A3 * 3 * eps(i) / (3 * eps(i) + 2 * EPS)

this mode is intended for the implicit membrane model(see below)

4 Same as CORR = 3 except that the local dielectric constant
is modulated spherically.

5 Same as CORR = 3 except that the local dielectric constant
is modulated cylindrically.

where r4 is volume integral over 1/r^4, r5 is square root
of integral over 1/r^5 and r7 is integral over 1/r^7 to the
power of 1/4.

TT Multiplicative factor for correction term (CORR = 0 only).

SHIFt The shifting factor of Alpha(i). MUST be set!

ESHIft Energy shifting factor of the self-polarization
energies: 1/Alpha(i). CORR=0 or 2 only. (Default 0.0)

SLOPE Multiplicative factor of the Alpha(i). CORR=1 or 2 only. (Default 1)

A1,A2, A3 Multiplicative factors in calculation of Alpha(i).

WATR The radius of the water probe. Usually this is set to 1.4
Angstroms. If this were changed, other parameters would have
to be modified.

EPSILON This is the value of the dielectric constant for the solvent medium.
The default value is 80.

KAPPA Debye-Huckel ionic term: Units of inverse length (Angs). Default
is 0 (no salt).

GEOM Select geometric cross-term in Still equation (default).

ARITH Select arithmetic cross-term in Still equation.

P6 Exponent in exponential of Still equation. Default is 4, for
historical reasons. Value of 8 is RECOMMENDED for GEOM, 6.5 for

WEIGHT Use WMAIN array for radii. (Default uses vdW radii array)

CLEAr Clear all arrays and logical flags used in Generalized Born
calculation. Use command by itself.

Parameters specific to GBMV I and II:

FIXA Update alphas only if coordinates have changed more
than expected for finite differences. Useful for
static pka calculations. With FIXA keyword, finite-difference
wouldn't work correctly, hence it must be specified. Not
on by default.

ALFRQ Update frequency of Born radii. Use with great caution!
One of LIMP,IMP, or EMP options must be selected. (Default 1)
Values other 1 not generally recommended.

LIMP Use ALFRQ*(dE/dalpha)(dalpha/dx) part of GB force every ALFRQ
steps. For ALFRQ <= 5.

EMP Decay constant of the impulse force. Default is 1.5, which
is meant for ALFRQ of 5. Generally, EMP ~= ALFRQ/4. For
ALFRQ <= 10. (Recommended option)

IMP Use (dE/dalpha)(dalpha/dx) part of GB force every ALFRQ
steps. Any ALFRQ can be used. Only meant for equilibrium

DN The cell width of the lookup grid. Larger values make program
slower. Smaller values use up more memory. Default of 1.0 A is best
compromise between speed and memory.

BETA Smoothing factor for tailing off of volume.
Values of around -100 are fine for GBMV I. Values between -8 to
-50 are reasonable for GBMV II. (Default -20)

Smaller values of beta lead to more stable dynamics, but compromise
the agreement with Poisson theory. In GBMV II the choice of BETA
also affects P3. Good pairs of values for GBMV II are:

BETA = -20, P3 = 0.70
BETA = -12, P3 = 0.65 * recommended as best compromise
BETA = -10, P3 = 0.57
BETA = -8, P3 = 0.35

LAMBda The threshold value for the atomic volumes. In GBMV I, smaller
values produces shorter Born radii and wide variance w/respect
to accurate PB radii. Large values produce larger radii but smaller
variance. In GBMV II, value should be kept at 0.5.

BUFR Distance that any atom is allowed to move before lookup table
is rebuilt. Larger values lead to less lookup table update but larger
memory usage. Use 0.0 for static structure.
Values between 0.2 and 1.0 Angstrom. (Default 0.5)

MEM Percentage extra memory beyond hypothetical calculation of table
size. (Default 10)

TOL Accuracy of the switching function used to determine accuracy of the
first derivatives, i.e. forces. (Default 1e-8)

SA Surface area coefficient (KCAL/(MOL*A**2)). (Default 0.0)
SASA Energy term shows up under EXTERN/ASP.

SB Surface area constant (KCAL/MOL) (no effect on forces) (Default 0.0)

SON The startpoint for the switching function of each hard sphere.
(Default 1.2) Units in Angstroms

SOFF The endpoint for the switching function of each hard sphere.
(Default 1.5)

P1 The multiplicative factor for the exponent of the
quartic exponential atomic function:

Gamma(i) = P1 * log(lambda)/(Rad(i)^4)

Parameters specific to GBMV II:

P1,P2 Variables which affect the shape of the VSA atomic function in the
region of R to R+2.

F(x) = A^2 / (A + x^2 - R^2)^2


A = P1 * R + P2 (Defaults: P1 = 1.25/P2 = 0.45)

P3 Scaling factor of VSA function. Default = 0.7
This factor depends on the value chosen for BETA
(see description above)

P4 Scaling coefficient for correction term to Still's equation.
(set to 0.0 for now)

P5 Exponent to the Still correction term. (use default for now)

HSX1/HSX2 Start and stop of hard-sphere tail with R(vdW) as origin.
(Defaults: -0.125/0.25).

ONX/OFFX Start and stop of VSA tail. Increasing values up to 2.8 A
makes better accuracy, however slows calculation. Compromise
of 1.9/2.1 is default.

FAST Turns on fast GBMV routine.

SGBFRQ Update frequency of internal lookup list in fast GBMV mode
(Default 1). Values between 1 and 10 are recommended.

SXD Delta used in fast GBMV mode lookup buffer. (Default 0).
Recommended values between 0.1 and 0.5. Requires 'FAST 1'

GBVDW If present, the VDW dispersion term is turned on.

GBASP If present, variable surface area coefficients are turned on.
They are read from the ASPValue array that should be set
before calling GBMV with appropriate scalar commands.

Parameters specific to HDGB (CORR = 3, 4, 5):

UNEPS Unit number of an input file holding the dielectric profile
values (Use -1 for the default profile). The format of this
input file is restricted. Comments are not allowed in a file.
The dielectric profile must be sampled in equal intervals.
The first line needs the number of sampling points n and the
sampling interval h (Angstrom). Two columns of the z coordinates
(Angstrom) and dielectric constants. The example is given
in test/data/hdgb_eps.dat.

A4,A5 Parameters in calculation of Alpha(i). A4 and A5 correspond
to the parameter D and E of Equation (15) in the reference (3)

ZS,ZM,ZT Parameters for a switching function for the nonpolar energy.
ST0 ZS, ZM, ZT, and ST0 corresponds to Za, Zb, Zc, and C of
Equation (11) in the reference (4).

UNNP Unit number of a previously opened (formatted) input file for
the non-polar profile. If this option is used the values given
with ZS, ZM, ZT, and ST0 are ignored. The format is the same
as the format used for the dielectric profile (see UNEPS).

HDGBRC If this flag is specified, the radius will be corrected
upon insertion to an implicit membrane.
(Only availabe for CORR = 3)

HDNOSW Turn off switching function for non-polar part when in HDGB
mode (CORR = 3,4,5)

EXCLGB Variable to exclude residues for HDGB to be used for inward facing
residue treatment. Used with atom selection.

Parameters specific to HDGBVDW (VDW dispersion term) (CORR = 3):

GBVD If present, the VDW dispersion term is turned on.

UNDO Unit number (81) of a previously opened density profile file for water
molecules. The format used for UNDO and other density profile files
(see below: UNDP, UNDC, UNDC2 and UNDN) is the same as the format
used for dielectric and non-polar profiles (see UNESP).

UNDP Unit number (82) of a previously opened density profile for oxygen
atoms of lipid molecules.

UNDC Unit number (83) of a previously opened density profiles for carbon
atoms at the head group of the lipid molecules.

UNDC2 Unit number (84) of a previously opened density profiles for carbon
atoms at the tail of the lipid molecules.

UNDN Unit number (85) of a previously opened density profiles for hydrogen
atoms of lipid molecules.

Parameters specific to DHDGB (CORR=3)
This is an extension of HDGB implicit membrane
with CORR=3 (planer membrane).
DHDGB will allow planer membranes local thickness to vary
dynamically. For more information about the DHDGB model please
refer to Panahi A., Feig M., JCTC 2013, 9, 1709-1719.
Please note that this model is only developed to be used with
single-pass transmembrane helices with radius of 7.5 A
For now, DHDGB only works for energy calculations (using energy
command), minimizations using SD and ABNR methods and VVER
integrator with Nose-Hoover thermostate (NVT ensemble)

QDHDGB keyword for turning membrane local fluctuation on
UNFHDGB unit number of a binary coefficient file that includes the
coefficients of spline interpolation of deformation energy.
Please use test/data/small_lookup_coeffs.bin as an example
for this unit. This file must meet specific criteria:
1. it has to be a binary file with size of real 4
2. it contains spline coefficients for deformation energy values
that exhustively sample between 35 to 0 A (membrane thickness)
with 5 A intervals
3. Since there are 7 intervals and each have one cubic term
(with 4 coefficients) the dimension of the number of coefficients
will be (4*7)^(ANGDIM) where ANGDIM can be either 3 or 5.

UNEPS_DEF_C, adjustable a{epsilon}, c{epsilon} and e{epsilon}
UNEPS_DEF_E of equation 24 above reference. The format of these files
are the same as UNEPS and UNNP in HDGB. To see an example
please see test/data/coeff_a_eps.txt,

UNNP_DEF_C, adjustable a{gamma}, c{gamma} and e{gamma}
UNNP_DEF_E of equation 27 above reference,The format of these files
are the same as UNEPS and UNNP in HDGB. To see an example
please see test/data/coeff_a_np.txt,

CIRCLERAD 7.5 radius of the inclusion cylinder
SMASS mass of virtual particles that represent the local position
membrane around the inclusion

ANGDIM 3 or 5 the number of S values on the contact circle.
The smallest value allowed is 3 and the biggest value is 5.
Please note that the file that contains the
cubic SPLINE coefficients must match the dimensions of
the s values (4*27)^(ANGDIM). We recommend using 5 S values for
each leaflet.

SDEF(1),... 25 initial position of the virtual particles

Please Note that during minimization the values of SDEF will be written in
units 120 for SD and 121 for ABNR respectively.
For now DHDGB is NOT compatible with GBMVFA option.

Parameters specific to Grid-based GBMV:

ML Number of surface points to carve out re-entrant surface

CONV Smear grid with cross-shaped blur function to improve accuracy

Additional parameters added for use with CPHMD:
HYBRID Keyword to invoke hybrid-solvent CPHMD. Allows GB radii
to be calculated be considering only a subset of the entire
system. ie. Ignoring solvent atoms
Used in conjuction with atom-selection

SELE Use the SELE keyword to manually specify atoms which will be
considered in the calculation of Born radii for use with CPHMD
*** note : GBMV does not currently support images,
therefore care should be used when attempting
to use GBMV with hybrid solvent PHMD using
periodic boundary conditions

Usage Examples and Compatibility

The examples below illustrate some of the uses of the generalized Born
Molecular Volume (GBMV) module. See c29test/gbmvtest.inp for more examples.

1) Coordinates MUST be defined for all atoms before invoking the GBMV keyword.
Otherwise, "infinite" grid is established which uses too much memory.

2) CUTOFF Parameters MUST be defined.
For non-infinite cutoffs, "switch" in nonbonded parameters is NECESSARY.

Example 1

!To perform a single-point energy calculation w/infinite cutoffs using
!GBMV I algorithm (any forcefield):

scalar wmain = radii

GBMV BETA -100 EPSILON 80 DN 1.0 WATR 1.4 TT 2.92 -
SHIFT -0.5 ESHIFT 0.0 LAMBDA1 0.1 P1 0.44 -
BUFR 0.5 Mem 20 CUTA 20 WTYP 0 -
WEIGHT ! Radii from wmain

ENERGY ctonnb 979 ctofnb 989 cutnb 999

Example 2

!To perform a single-point energy calculation w/infinite cutoffs using
!the GBMV II algorithm (any forcefield):

GBMV BETA -20 EPSILON 80 DN 1.0 watr 1.4 GEOM -
TOL 1e-8 BUFR 0.5 Mem 10 CUTA 20 HSX1 -0.125 HSX2 0.25 -
ALFRQ 1 EMP 1.5 P4 0.0 P6 8.0 P3 0.70 ONX 1.9 OFFX 2.1 -
WTYP 2 NPHI 38 SHIFT -0.102 SLOPE 0.9085 CORR 1

ENERGY ctonnb 979 ctofnb 989 cutnb 999

GBMV CLEAR ! Clear GB arrays

Example 3

!Recommended setup for molecular dynamics simulations with
!the GBMV II algorithm:

UPDATE atom CDIE eps 1 cutnb 21 ctofnb 18 ctonnb 16 switch vswitch

GEOM BETA -12 P1 0.45 P2 1.25 P3 0.65 P6 8.0 -

!You should use Langevin dynamics and a 1.5 fs time step (with SHAKE)
!is recommended for optimal stability. Many applications will also
!tolerate 2 fs time step
!(more info in: Chocholousova & Feig, JCC (2006) 27, 719-729)




Example 5

!Grid-based GBMV:

GBMV GRID EPSILON 80 DN 0.2 watr 1.4 GEOM P6 8.0 -
WTYP 0 NPHI 10 SHIFT -0.007998 SLOPE 0.9026 CORR 1 CONV

ENERGY ctonnb 979 ctofnb 989 cutnb 999

Example 6

! HDGB DPPC membrane
! If you want the default DPPC profile used in the reference (4),
! comment out the open file statement and set UNEPS to -1.

! The input file will be closed automatically, so you don't need
! the explicit close statement.
open unit 1 name eps.dat read form

GBMV A1 0.3255 A3 1.085 A4 -0.14 A5 -0.15 -
ZS 0.5 ZM 9.2 ZT 25 ST0 0.32

ENERGY ctonnb 979 ctofnb 989 cutnb 999

Example 7

! HDGBVDW DPPC membrane

open unit 89 name hdgbvdw_eps.dat read form
open unit 88 name hdgbvdw_np.dat read form

open unit 81 name hdgbvdw_water.txt read form
open unit 82 name hdgbvdw_o.txt read form
open unit 83 name hdgbvdw_c.txt read form
open unit 84 name hdgbvdw_cc.txt read form
open unit 85 name hdgbvdw_h.txt read form

CORR 3 -
UNEPS 89 UNNP 88 -
UNDO 81 UNDP 82 UNDC 83 UND2C 84 UNDN 85 -
SA 0.038 SB 0

ENERGY ctonnb 979 ctofnb 989 cutnb 999

<Known Compatible with>
- VIBRAN (finite difference second derivatives)
- MMFF (WEIGHT keyword must be used)

<Known Incompatible with (so far)>
- VIBRAN (no analytic second derivatives)
- multiple dielectric
- QUANTUM* (single energy with original charges is ok)