******************************************************************************** 7.3.4 Comparing with Delphi results Under identical condition, PBSA is highly consistent with Delphi in term of computed reaction field energies. In this subsection, we briefly go over the details on how you can obtain comparable energies from both program. Apparently, you need coordinates, atomic charges, and atomic radii that have exactly the same numerical values but in both the Amber format and the Delphi format, i.e. the pqr format. For a Delphi computation with the following input parameters: salt=0.150 ionrad=2.0 exdi=80.0 indi=1.0 scale=2.0 prbrad=1.5 perfil=50 bndcon=4 linit=1000 a comparable computation in PBSA can be obtained by using the following input file: Sample PB for delphi comparison &cntrl ntx=1, imin=1, ipb=1, inp=0 / &pb npbverb=0, istrng=150, epsout=80.0, epsin=1.0, ivalence=1, iprob=2.0, space=0.5, accept=1e-3, dprob=1.5, radiopt=0, fillratio=2, bcopt=6, smoothopt=2, nfocus=1, dbfopt=1, cutnb=0, maxitn=10000 / The above sample input files are also provided in the current release under $AMBERHOME/test/pbsa_delphi. Note that the values of exdi, indi, prbrad, and ionrad in Delphi should be consistent with the values of epsout, epsin, dprob, and iprob in PBSA, respectively. In Delphi salt=0.150 is set in the unit of M, while in PBSA istrng=150 is in the unit of mM. In Delphi the grid spacing is set as the number of grids per \AAngstrom, i.e. scale=2.0, while in PBSA the grid spacing is set straight as space=0.5 in \AAngstrom. In Delphi the grid dimension is set as percentage of the solute dimension over the grid dimension, i.e. perfil=50, which is equivalent to the ratio of solute dimension over grid dimension set as fillratio = 2 in PBSA. Finally, Delphi sets the boundary condition by bndcnd=4 and PBSA sets the boundary condition as bcopt=6; both programs mean to use the Debye-HŸckel limitation behavior for each atomic charged sphere. There are additional options in PBSA that do not have corresponding counterparts in Delphi. For example, smoothopt is used to instruct the program to use a specific dielectric boundary smoothing option, which is equivalent to that used in Delphi when set to 2. (see Section 7.1.3); arcres is used to set the resolution of numerical dot representation of the solvent accessible arcs, the default is 1/16 \AAngstrom, which is similar to the numerical resolution of solvent accessible arc dots used in Delphi. 7.4 PBSA in SANDER All PBSA functionalities are available in SANDER and all input options are exactly the same as in the standalone PBSA. Apparent exceptions are ipb and inp: you need to really set ipb/inp to nonzero in order to invoke PBSA functionalities. 7.5 PBSA in NAB PBSA functionalities are available in NAB as a part of the standard build. However the available input options are limited, please refer to the table in Section 14.1 for the list of available PBSA input options. The structures and parameters are supplied by NAB's facility. An extra feature for PBSA in NAB is the availability of electrostatic forces or gradients. We describe special consideration in its applications in the following section. 7.5.1 Electrostatic Forces/Gradients in PBSA Force calculation in the finite-difference Poisson-Boltzmann method is straightforward, though not a trivial issue. It can be shown, by using the variation of the electrostatic free energy, that the electrostatic force density consists of three components, viz., the reaction field force, the dielectric boundary force, and the ionic force. \begin_inset CommandInset citation LatexCommand cite key "Gilson:1993p11929" \end_inset Since the ionic force is much smaller in absolute value than the other two components, we only include the reaction field force and the dielectric boundary force in this release. The reaction field force only exists where there are atomic charges, so that it is straightforward to be mapped onto atoms. In contrast, the dielectric boundary force exists on the molecular surface where the dielectric constant changes. The surface force, or pressure, cannot be easily mapped onto atoms. This is because a force-mapping procedure from the molecular surface to atoms apparently needs the derivatives of molecular surface with respect to atomic positions. However such derivatives do not exist for the widely used molecular surface definition, i.e. the solvent excluded surface (SES). We are actively developing an analytical molecular surface definition that is consistent with the widely used SES definition for the numerical PB methods so that this difficulty will be overcome in future releases. Temporarily, there is a partial solution in the mapping of dielectric boundary force as described by Gilson et al \begin_inset CommandInset citation LatexCommand cite key "Gilson:1993p11929" \end_inset when the SES definition is used. This is currently adopted in the PBSA module. Note that the method is rigorously correct only for two overlapping atoms. On solvent-exposed surface with well-packed and highly charged atoms, such as nucleic acid backbones, the error in the force-mapping procedure can be very large. Our tests show that the deviations from the numerical finite-difference forces can be in the orders of several tenths of kcal/mol-\AAngstrom. In other situations, the errors are in the orders of several hundredths of kcal/mol-\AAngstrom. 7.5.2 Example . . . mm_options("ntpr=1, cut=99.0"); // No solute-solute cutoff mm_options("ipb=1"); // Use PBSA mm_options("accept=0.000001"); // Convergence criterion mm_options("sprob=1.6"); // Solvent probe radius for SASA mm_options("radiopt=1"); // Useatom-type/charge-based radii mm_options("fillratio=4"); // Coarse/Fine ratio of electrostatic focusing . . . ******************************************************************************** @Article{Gilson:1993p11929, author = {Gilson, M.K. and Davis, M.E and Luty, B.A. and McCammon, J.A.}, journal = {J Phys Chem}, title = {{Computation of electrostatic forces on solvated molecules using the Poisson-Boltzmann equation}}, number = {14}, pages = {3591--3600}, volume = {97}, year = {1993} }