## apbs-users

 [Apbs-users] help for a beginner: any serious problems with this input file? From: Daniel Ensign - 2010-07-01 19:49:10 ```Dear APBS users list, I am trying to calculate the electric field along a certain bond at the interface of protein and water (and later, protein+protein). However, I've had a hard time so far getting the results to compare well with our experimental observables, and I'm worried that I'm not giving PB the best possible chance to get the "right" answer. (Whether the experimental interpretation is correct is another matter, of course.) Would it be possible to get some criticism (constructive or destructive, either way) on the settings in this APBS input file? Am I making any horrible mistakes? Thanks, Dan Ensign read mol pqr frame.pqr end elec mg-auto dime 193 193 193 cglen 74 74 79 fglen 63 64 60 fgcent mol 1 cgcent mol 1 mol 1 npbe bcfl mdh ion charge 1 conc 0.075 radius 2.0 ion charge -1 conc 0.075 radius 2.0 pdie 2.0 sdie 78.0 chgm spl4 srfm mol srad 1.4 # does not matter swin 0.3 # irrelevant with srfm mol sdens 10.0 # apbs suggested value temp 300.0 calcenergy total calcforce no write pot dx pot end quit -- Tu ne cede malis, sed contra audentior ito. -- Virgil ```
 Re: [Apbs-users] help for a beginner: any serious problems with this input file? From: Nathan Baker - 2010-07-02 02:15:42 ```Hi Dan -- For the application you're describing, you need to break the calculation into two pieces. You should be using the Poisson-Boltzmann equation to calculate the reaction field the potential: the difference between the potential in the inhomogeneous dielectric (as you have below) and the potential in a homogeneous dielectric (with both pdie and side set to 2.0). This should then be added to a Coulombic potential calculation (with a dielectric of 2). The reason for this decomposition is numerical accuracy: the numerical Poisson-Boltzmann solvers can calculate the slowly-varying reaction field potential with reasonable accuracy but have a hard time resolving the more-quickly-varying Coulombic potential with any level of accuracy -- especially near the charge centers you're interested in. Hope this helps! -- Nathan On Thu, Jul 1, 2010 at 12:15 PM, Daniel Ensign wrote: > Dear APBS users list, > > I am trying to calculate the electric field along a certain bond at > the interface of protein and water (and later, protein+protein). > However, I've had a hard time so far getting the results to compare well > with our experimental observables, and I'm worried that I'm not giving > PB the best possible chance to get the "right" answer. > > (Whether the experimental interpretation is correct is another matter, > of course.) > > Would it be possible to get some criticism (constructive or > destructive, either way) on the settings in this APBS input file? Am I > making any horrible mistakes? > > Thanks, > Dan Ensign > > read >    mol pqr frame.pqr > end > > elec >    mg-auto >    dime 193 193 193 >    cglen 74 74 79 >    fglen 63 64 60 >    fgcent mol 1 >    cgcent mol 1 >    mol 1 >    npbe >    bcfl mdh >    ion charge 1 conc 0.075 radius 2.0 >    ion charge -1 conc 0.075 radius 2.0 >    pdie 2.0 >    sdie 78.0 >    chgm spl4 >    srfm mol >    srad 1.4 # does not matter >    swin 0.3 # irrelevant with srfm mol >    sdens 10.0 # apbs suggested value >    temp 300.0 >    calcenergy total >    calcforce no >    write pot dx pot > end > > > quit > > > -- > Tu ne cede malis, sed contra audentior ito. -- Virgil > > ------------------------------------------------------------------------------ > This SF.net email is sponsored by Sprint > What will you do first with EVO, the first 4G phone? > Visit sprint.com/first -- http://p.sf.net/sfu/sprint-com-first > _______________________________________________ > apbs-users mailing list > apbs-users@... > https://lists.sourceforge.net/lists/listinfo/apbs-users > ```
 Re: [Apbs-users] help for a beginner: any serious problems with this input file? From: Daniel Ensign - 2010-07-05 20:56:18 ```Hi Prof. Baker, thanks for the reply, Two more questions: 1. You suggest calculating "the reaction field [of] the potential" for reasons of numerical accuracy. That's good, but regarding that please see question 2. But does this mean that the calculation I'm proposing is not likely to be any good at all? Several of the settings are making me sweat a little: for instance, I was using "chgm spl0" in a previous calc and after more reading I don't think that's too smart for me. We've had a poor correlation of those calculated results with our experimental data (please email Prof. Webb here at The University of Texas; I'm sure she'd send you a manuscript draft if you are interested) and I would like to avoid the reason for that poor correlation being that I did the calculation totally wrong. One way we're approaching this is to say, how would random biochemist X use PB, as a beginner, to calculate the fields that we're measuring? That way if the calculation fails, we at least show that one can be extremely naive using PB. This approach has the additional benefit that I actually am a PB beginner. However, as the computational guy in all of this, I need to give the calculation the best shot possible, just to show my experimental labmates who's really got it together. :) On the other hand, maybe your suggestions about numerical accuracy and the reaction field indicate that I should expect a poor correlation for calculations which don't consider the RF, no matter how perfect my APBS settings could be, which brings me to ... 2. I'm not sure I fully understand the reaction field + pure Coulomb calculation, so may I attempt to state what it is that I think I'm reading and have your corrections of my errors? First, I calculate the reaction field by doing a calculation with values of pdie=pdie1 and sdie=sdie1. Then I repeat the same calculation but with pdie=pdie1 and sdie=pdie1 (the same, small value). I subtract the potential of the first calculation from the potential of the second. Then, I calculate the full Coulombic interaction between all of the protein atoms in my system, based on the same point-charge force field I used for sampling. (Can APBS do this easily for me?) I add that potential to the first one. This gives me the potential I want. Is that correct? Thanks for your time, Dan On Thu, 1 Jul 2010 19:15:31 -0700 Nathan Baker wrote: > Hi Dan -- > > For the application you're describing, you need to break the > calculation into two pieces. > > You should be using the Poisson-Boltzmann equation to calculate the > reaction field the potential: the difference between the potential in > the inhomogeneous dielectric (as you have below) and the potential in > a homogeneous dielectric (with both pdie and side set to 2.0). This > should then be added to a Coulombic potential calculation (with a > dielectric of 2). > > The reason for this decomposition is numerical accuracy: the > numerical Poisson-Boltzmann solvers can calculate the slowly-varying > reaction field potential with reasonable accuracy but have a hard time > resolving the more-quickly-varying Coulombic potential with any level > of accuracy -- especially near the charge centers you're interested > in. > > Hope this helps! > > -- Nathan > > On Thu, Jul 1, 2010 at 12:15 PM, Daniel Ensign > wrote: > > Dear APBS users list, > > > > I am trying to calculate the electric field along a certain bond at > > the interface of protein and water (and later, protein+protein). > > However, I've had a hard time so far getting the results to compare > > well with our experimental observables, and I'm worried that I'm > > not giving PB the best possible chance to get the "right" answer. > > > > (Whether the experimental interpretation is correct is another > > matter, of course.) > > > > Would it be possible to get some criticism (constructive or > > destructive, either way) on the settings in this APBS input file? > > Am I making any horrible mistakes? > > > > Thanks, > > Dan Ensign > > > > read > >    mol pqr frame.pqr > > end > > > > elec > >    mg-auto > >    dime 193 193 193 > >    cglen 74 74 79 > >    fglen 63 64 60 > >    fgcent mol 1 > >    cgcent mol 1 > >    mol 1 > >    npbe > >    bcfl mdh > >    ion charge 1 conc 0.075 radius 2.0 > >    ion charge -1 conc 0.075 radius 2.0 > >    pdie 2.0 > >    sdie 78.0 > >    chgm spl4 > >    srfm mol > >    srad 1.4 # does not matter > >    swin 0.3 # irrelevant with srfm mol > >    sdens 10.0 # apbs suggested value > >    temp 300.0 > >    calcenergy total > >    calcforce no > >    write pot dx pot > > end > > > > > > quit > > > > > > -- > > Tu ne cede malis, sed contra audentior ito. -- Virgil > > > > ------------------------------------------------------------------------------ > > This SF.net email is sponsored by Sprint > > What will you do first with EVO, the first 4G phone? > > Visit sprint.com/first -- http://p.sf.net/sfu/sprint-com-first > > _______________________________________________ > > apbs-users mailing list > > apbs-users@... > > https://lists.sourceforge.net/lists/listinfo/apbs-users > > > > ------------------------------------------------------------------------------ > This SF.net email is sponsored by Sprint > What will you do first with EVO, the first 4G phone? > Visit sprint.com/first -- http://p.sf.net/sfu/sprint-com-first > _______________________________________________ > apbs-users mailing list > apbs-users@... > https://lists.sourceforge.net/lists/listinfo/apbs-users -- Tu ne cede malis, sed contra audentior ito. -- Virgil ```
 Re: [Apbs-users] help for a beginner: any serious problems with this input file? From: Baker, Nathan - 2010-07-06 13:46:15 ```Hi Dan -- > 1. You suggest calculating "the reaction field [of] the potential" for > reasons of numerical accuracy. That's good, but regarding that please > see question 2. But does this mean that the calculation I'm proposing > is > not likely to be any good at all? [NAB] The calculation represented by the input file you attached before is not a good idea due to the inability of most finite difference and finite element codes (solving the traditional form of the Poisson-Boltzmann equation) to accurately resolve the very large potentials near charge centers. > Several of the settings are making me > sweat a little: for instance, I was using "chgm spl0" in a previous > calc and after more reading I don't think that's too smart for me. [NAB] That's the most localized version of the charge distribution -- but also the one that's most sensitive to grid setup. > We've > had a poor correlation of those calculated results with our > experimental > data (please email Prof. Webb here at The University of Texas; I'm sure > she'd send you a manuscript draft if you are interested) and I would > like to avoid the reason for that poor correlation being that I did the > calculation totally wrong. [NAB] :) > One way we're approaching this is to say, how would random biochemist X > use PB, as a beginner, to calculate the fields that we're measuring? > That way if the calculation fails, we at least show that one can be > extremely naive using PB. This approach has the additional benefit that > I actually am a PB beginner. However, as the computational guy in all > of this, I need to give the calculation the best shot possible, just to > show my experimental labmates who's really got it together. :) > > On the other hand, maybe your suggestions about numerical accuracy and > the reaction field indicate that I should expect a poor correlation > for calculations which don't consider the RF, no matter how perfect my > APBS settings could be, which brings me to ... > > 2. I'm not sure I fully understand the reaction field + pure Coulomb > calculation, so may I attempt to state what it is that I think I'm > reading and have your corrections of my errors? > > First, I calculate the reaction field by doing a calculation with > values of pdie=pdie1 and sdie=sdie1. Then I repeat the same calculation > but with pdie=pdie1 and sdie=pdie1 (the same, small value). I subtract > the potential of the first calculation from the potential of the > second. [NAB] Yes, that's correct. > Then, I calculate the full Coulombic interaction between all of the > protein atoms in my system, based on the same point-charge force > field I used for sampling. (Can APBS do this easily for me?) [NAB] Yes, APBS provides the tool "coulomb" for this. Note that it calculates the potential for a vacuum, so you may need to rescale to pdie1. Also note that it's important to use the same charge distribution for all steps. > I add that potential to the first one. This gives me the potential I > want. [NAB] You got it! -- Nathan > Is that correct? > > Thanks for your time, > Dan > > On Thu, 1 Jul 2010 19:15:31 -0700 > Nathan Baker wrote: > > > Hi Dan -- > > > > For the application you're describing, you need to break the > > calculation into two pieces. > > > > You should be using the Poisson-Boltzmann equation to calculate the > > reaction field the potential: the difference between the potential > in > > the inhomogeneous dielectric (as you have below) and the potential in > > a homogeneous dielectric (with both pdie and side set to 2.0). This > > should then be added to a Coulombic potential calculation (with a > > dielectric of 2). > > > > The reason for this decomposition is numerical accuracy: the > > numerical Poisson-Boltzmann solvers can calculate the slowly-varying > > reaction field potential with reasonable accuracy but have a hard > time > > resolving the more-quickly-varying Coulombic potential with any level > > of accuracy -- especially near the charge centers you're interested > > in. > > > > Hope this helps! > > > > -- Nathan > > > > On Thu, Jul 1, 2010 at 12:15 PM, Daniel Ensign > > wrote: > > > Dear APBS users list, > > > > > > I am trying to calculate the electric field along a certain bond at > > > the interface of protein and water (and later, protein+protein). > > > However, I've had a hard time so far getting the results to compare > > > well with our experimental observables, and I'm worried that I'm > > > not giving PB the best possible chance to get the "right" answer. > > > > > > (Whether the experimental interpretation is correct is another > > > matter, of course.) > > > > > > Would it be possible to get some criticism (constructive or > > > destructive, either way) on the settings in this APBS input file? > > > Am I making any horrible mistakes? > > > > > > Thanks, > > > Dan Ensign > > > > > > read > > >    mol pqr frame.pqr > > > end > > > > > > elec > > >    mg-auto > > >    dime 193 193 193 > > >    cglen 74 74 79 > > >    fglen 63 64 60 > > >    fgcent mol 1 > > >    cgcent mol 1 > > >    mol 1 > > >    npbe > > >    bcfl mdh > > >    ion charge 1 conc 0.075 radius 2.0 > > >    ion charge -1 conc 0.075 radius 2.0 > > >    pdie 2.0 > > >    sdie 78.0 > > >    chgm spl4 > > >    srfm mol > > >    srad 1.4 # does not matter > > >    swin 0.3 # irrelevant with srfm mol > > >    sdens 10.0 # apbs suggested value > > >    temp 300.0 > > >    calcenergy total > > >    calcforce no > > >    write pot dx pot > > > end > > > > > > > > > quit > > > > > > > > > -- > > > Tu ne cede malis, sed contra audentior ito. -- Virgil > > > > > > ------------------------------------------------------------------- > ----------- > > > This SF.net email is sponsored by Sprint > > > What will you do first with EVO, the first 4G phone? > > > Visit sprint.com/first -- http://p.sf.net/sfu/sprint-com-first > > > _______________________________________________ > > > apbs-users mailing list > > > apbs-users@... > > > https://lists.sourceforge.net/lists/listinfo/apbs-users > > > > > > > --------------------------------------------------------------------- > --------- > > This SF.net email is sponsored by Sprint > > What will you do first with EVO, the first 4G phone? > > Visit sprint.com/first -- http://p.sf.net/sfu/sprint-com-first > > _______________________________________________ > > apbs-users mailing list > > apbs-users@... > > https://lists.sourceforge.net/lists/listinfo/apbs-users > > > > -- > Tu ne cede malis, sed contra audentior ito. -- Virgil > > ----------------------------------------------------------------------- > ------- > This SF.net email is sponsored by Sprint > What will you do first with EVO, the first 4G phone? > Visit sprint.com/first -- http://p.sf.net/sfu/sprint-com-first > _______________________________________________ > apbs-users mailing list > apbs-users@... > https://lists.sourceforge.net/lists/listinfo/apbs-users ```
 Re: [Apbs-users] help for a beginner: any serious problems with this input file? From: Daniel Ensign - 2010-07-07 15:40:46 ```Professor Baker, Another question (watch, I'm full of 'em): when you say "large potentials near charge centers," what do you mean by "large" and "near"? One way we're doing our calculation, since we're interested in *changes* of the potential due to mutation, etc., rather than the potential itself, is to ignore the charges of atoms near the point of interest, which are constant across all of our systems. (In other words, we set those charges to zero in the force field.) I wonder if this is far enough away (at least ~1.5 C-C bond lengths) from charge centers to yield sensible results. Dan On Tue, 6 Jul 2010 06:45:59 -0700 "Baker, Nathan" wrote: > Hi Dan -- > > > 1. You suggest calculating "the reaction field [of] the potential" > > for reasons of numerical accuracy. That's good, but regarding that > > please see question 2. But does this mean that the calculation I'm > > proposing is > > not likely to be any good at all? > > [NAB] The calculation represented by the input file you attached > before is not a good idea due to the inability of most finite > difference and finite element codes (solving the traditional form of > the Poisson-Boltzmann equation) to accurately resolve the very large > potentials near charge centers. > > > Several of the settings are making me > > sweat a little: for instance, I was using "chgm spl0" in a previous > > calc and after more reading I don't think that's too smart for me. > > [NAB] That's the most localized version of the charge distribution -- > but also the one that's most sensitive to grid setup. > > > We've > > had a poor correlation of those calculated results with our > > experimental > > data (please email Prof. Webb here at The University of Texas; I'm > > sure she'd send you a manuscript draft if you are interested) and I > > would like to avoid the reason for that poor correlation being that > > I did the calculation totally wrong. > > [NAB] :) > > > One way we're approaching this is to say, how would random > > biochemist X use PB, as a beginner, to calculate the fields that > > we're measuring? That way if the calculation fails, we at least > > show that one can be extremely naive using PB. This approach has > > the additional benefit that I actually am a PB beginner. However, > > as the computational guy in all of this, I need to give the > > calculation the best shot possible, just to show my experimental > > labmates who's really got it together. :) > > > > On the other hand, maybe your suggestions about numerical accuracy > > and the reaction field indicate that I should expect a poor > > correlation for calculations which don't consider the RF, no matter > > how perfect my APBS settings could be, which brings me to ... > > > > 2. I'm not sure I fully understand the reaction field + pure Coulomb > > calculation, so may I attempt to state what it is that I think I'm > > reading and have your corrections of my errors? > > > > First, I calculate the reaction field by doing a calculation with > > values of pdie=pdie1 and sdie=sdie1. Then I repeat the same > > calculation but with pdie=pdie1 and sdie=pdie1 (the same, small > > value). I subtract the potential of the first calculation from the > > potential of the second. > > [NAB] Yes, that's correct. > > > Then, I calculate the full Coulombic interaction between all of the > > protein atoms in my system, based on the same point-charge force > > field I used for sampling. (Can APBS do this easily for me?) > > [NAB] Yes, APBS provides the tool "coulomb" for this. Note that it > calculates the potential for a vacuum, so you may need to rescale to > pdie1. Also note that it's important to use the same charge > distribution for all steps. > > I add that potential to the first one. This gives me the potential I > > want. > > [NAB] You got it! > > -- Nathan > > > Is that correct? > > > > Thanks for your time, > > Dan > > > > On Thu, 1 Jul 2010 19:15:31 -0700 > > Nathan Baker wrote: > > > > > Hi Dan -- > > > > > > For the application you're describing, you need to break the > > > calculation into two pieces. > > > > > > You should be using the Poisson-Boltzmann equation to calculate > > > the reaction field the potential: the difference between the > > > potential > > in > > > the inhomogeneous dielectric (as you have below) and the > > > potential in a homogeneous dielectric (with both pdie and side > > > set to 2.0). This should then be added to a Coulombic potential > > > calculation (with a dielectric of 2). > > > > > > The reason for this decomposition is numerical accuracy: the > > > numerical Poisson-Boltzmann solvers can calculate the > > > slowly-varying reaction field potential with reasonable accuracy > > > but have a hard > > time > > > resolving the more-quickly-varying Coulombic potential with any > > > level of accuracy -- especially near the charge centers you're > > > interested in. > > > > > > Hope this helps! > > > > > > -- Nathan > > > > > > On Thu, Jul 1, 2010 at 12:15 PM, Daniel Ensign > > > wrote: > > > > Dear APBS users list, > > > > > > > > I am trying to calculate the electric field along a certain > > > > bond at the interface of protein and water (and later, > > > > protein+protein). However, I've had a hard time so far getting > > > > the results to compare well with our experimental observables, > > > > and I'm worried that I'm not giving PB the best possible chance > > > > to get the "right" answer. > > > > > > > > (Whether the experimental interpretation is correct is another > > > > matter, of course.) > > > > > > > > Would it be possible to get some criticism (constructive or > > > > destructive, either way) on the settings in this APBS input > > > > file? Am I making any horrible mistakes? > > > > > > > > Thanks, > > > > Dan Ensign > > > > > > > > read > > > >    mol pqr frame.pqr > > > > end > > > > > > > > elec > > > >    mg-auto > > > >    dime 193 193 193 > > > >    cglen 74 74 79 > > > >    fglen 63 64 60 > > > >    fgcent mol 1 > > > >    cgcent mol 1 > > > >    mol 1 > > > >    npbe > > > >    bcfl mdh > > > >    ion charge 1 conc 0.075 radius 2.0 > > > >    ion charge -1 conc 0.075 radius 2.0 > > > >    pdie 2.0 > > > >    sdie 78.0 > > > >    chgm spl4 > > > >    srfm mol > > > >    srad 1.4 # does not matter > > > >    swin 0.3 # irrelevant with srfm mol > > > >    sdens 10.0 # apbs suggested value > > > >    temp 300.0 > > > >    calcenergy total > > > >    calcforce no > > > >    write pot dx pot > > > > end > > > > > > > > > > > > quit > > > > > > > > > > > > -- > > > > Tu ne cede malis, sed contra audentior ito. -- Virgil > > > > > > > > ------------------------------------------------------------------- > > ----------- > > > > This SF.net email is sponsored by Sprint > > > > What will you do first with EVO, the first 4G phone? > > > > Visit sprint.com/first -- http://p.sf.net/sfu/sprint-com-first > > > > _______________________________________________ > > > > apbs-users mailing list > > > > apbs-users@... > > > > https://lists.sourceforge.net/lists/listinfo/apbs-users > > > > > > > > > > --------------------------------------------------------------------- > > --------- > > > This SF.net email is sponsored by Sprint > > > What will you do first with EVO, the first 4G phone? > > > Visit sprint.com/first -- http://p.sf.net/sfu/sprint-com-first > > > _______________________________________________ > > > apbs-users mailing list > > > apbs-users@... > > > https://lists.sourceforge.net/lists/listinfo/apbs-users > > > > > > > > -- > > Tu ne cede malis, sed contra audentior ito. -- Virgil > > > > ----------------------------------------------------------------------- > > ------- > > This SF.net email is sponsored by Sprint > > What will you do first with EVO, the first 4G phone? > > Visit sprint.com/first -- http://p.sf.net/sfu/sprint-com-first > > _______________________________________________ > > apbs-users mailing list > > apbs-users@... > > https://lists.sourceforge.net/lists/listinfo/apbs-users -- Tu ne cede malis, sed contra audentior ito. -- Virgil ```
 Re: [Apbs-users] help for a beginner: any serious problems with this input file? From: Baker, Nathan - 2010-07-07 16:08:06 ```Hi Dan -- > Another question (watch, I'm full of 'em): when you say "large > potentials near charge centers," what do you mean by "large" and > "near"?' [NAB] I'm basically referring to the behavior of Coulomb's Law near these charge centers. Finite difference methods will not accurately resolve the rapidly-changing potential near the charge singularities. "Near" is really a function of grid points -- your accuracy will deteriorate as you reach distances with fewer grid points between the point of observation and the charge location. "Large" could be infinite -- for an infinitely fine grid. In practice, there is an upper limit on the potential value at the charge center that is related to the grid spacing. > One way we're doing our calculation, since we're interested > in *changes* of the potential due to mutation, etc., rather than the > potential itself, is to ignore the charges of atoms near the point of > interest, which are constant across all of our systems. (In other > words, we set those charges to zero in the force field.) > > I wonder if this is far enough away (at least ~1.5 C-C bond lengths) > from charge centers to yield sensible results. [NAB] I would examine the sensitivity of your results as a function of grid dimensions (e.g., "dime") for a fixed set of grid lengths. Ideally, you would have several grid points between the charge location and where you're measuring the potential. Note that if you're measuring the field, you'll need even higher resolution for the same level of accuracy. Thanks, Nathan > Dan > > On Tue, 6 Jul 2010 06:45:59 -0700 > "Baker, Nathan" wrote: > > > Hi Dan -- > > > > > 1. You suggest calculating "the reaction field [of] the potential" > > > for reasons of numerical accuracy. That's good, but regarding that > > > please see question 2. But does this mean that the calculation I'm > > > proposing is > > > not likely to be any good at all? > > > > [NAB] The calculation represented by the input file you attached > > before is not a good idea due to the inability of most finite > > difference and finite element codes (solving the traditional form of > > the Poisson-Boltzmann equation) to accurately resolve the very large > > potentials near charge centers. > > > > > Several of the settings are making me > > > sweat a little: for instance, I was using "chgm spl0" in a previous > > > calc and after more reading I don't think that's too smart for me. > > > > [NAB] That's the most localized version of the charge distribution -- > > but also the one that's most sensitive to grid setup. > > > > > We've > > > had a poor correlation of those calculated results with our > > > experimental > > > data (please email Prof. Webb here at The University of Texas; I'm > > > sure she'd send you a manuscript draft if you are interested) and I > > > would like to avoid the reason for that poor correlation being that > > > I did the calculation totally wrong. > > > > [NAB] :) > > > > > One way we're approaching this is to say, how would random > > > biochemist X use PB, as a beginner, to calculate the fields that > > > we're measuring? That way if the calculation fails, we at least > > > show that one can be extremely naive using PB. This approach has > > > the additional benefit that I actually am a PB beginner. However, > > > as the computational guy in all of this, I need to give the > > > calculation the best shot possible, just to show my experimental > > > labmates who's really got it together. :) > > > > > > On the other hand, maybe your suggestions about numerical accuracy > > > and the reaction field indicate that I should expect a poor > > > correlation for calculations which don't consider the RF, no matter > > > how perfect my APBS settings could be, which brings me to ... > > > > > > 2. I'm not sure I fully understand the reaction field + pure > Coulomb > > > calculation, so may I attempt to state what it is that I think I'm > > > reading and have your corrections of my errors? > > > > > > First, I calculate the reaction field by doing a calculation with > > > values of pdie=pdie1 and sdie=sdie1. Then I repeat the same > > > calculation but with pdie=pdie1 and sdie=pdie1 (the same, small > > > value). I subtract the potential of the first calculation from the > > > potential of the second. > > > > [NAB] Yes, that's correct. > > > > > Then, I calculate the full Coulombic interaction between all of the > > > protein atoms in my system, based on the same point-charge force > > > field I used for sampling. (Can APBS do this easily for me?) > > > > [NAB] Yes, APBS provides the tool "coulomb" for this. Note that it > > calculates the potential for a vacuum, so you may need to rescale to > > pdie1. Also note that it's important to use the same charge > > distribution for all steps. > > > I add that potential to the first one. This gives me the potential > I > > > want. > > > > [NAB] You got it! > > > > -- Nathan > > > > > Is that correct? > > > > > > Thanks for your time, > > > Dan > > > > > > On Thu, 1 Jul 2010 19:15:31 -0700 > > > Nathan Baker wrote: > > > > > > > Hi Dan -- > > > > > > > > For the application you're describing, you need to break the > > > > calculation into two pieces. > > > > > > > > You should be using the Poisson-Boltzmann equation to calculate > > > > the reaction field the potential: the difference between the > > > > potential > > > in > > > > the inhomogeneous dielectric (as you have below) and the > > > > potential in a homogeneous dielectric (with both pdie and side > > > > set to 2.0). This should then be added to a Coulombic potential > > > > calculation (with a dielectric of 2). > > > > > > > > The reason for this decomposition is numerical accuracy: the > > > > numerical Poisson-Boltzmann solvers can calculate the > > > > slowly-varying reaction field potential with reasonable accuracy > > > > but have a hard > > > time > > > > resolving the more-quickly-varying Coulombic potential with any > > > > level of accuracy -- especially near the charge centers you're > > > > interested in. > > > > > > > > Hope this helps! > > > > > > > > -- Nathan > > > > > > > > On Thu, Jul 1, 2010 at 12:15 PM, Daniel Ensign > > > > wrote: > > > > > Dear APBS users list, > > > > > > > > > > I am trying to calculate the electric field along a certain > > > > > bond at the interface of protein and water (and later, > > > > > protein+protein). However, I've had a hard time so far getting > > > > > the results to compare well with our experimental observables, > > > > > and I'm worried that I'm not giving PB the best possible chance > > > > > to get the "right" answer. > > > > > > > > > > (Whether the experimental interpretation is correct is another > > > > > matter, of course.) > > > > > > > > > > Would it be possible to get some criticism (constructive or > > > > > destructive, either way) on the settings in this APBS input > > > > > file? Am I making any horrible mistakes? > > > > > > > > > > Thanks, > > > > > Dan Ensign > > > > > > > > > > read > > > > >    mol pqr frame.pqr > > > > > end > > > > > > > > > > elec > > > > >    mg-auto > > > > >    dime 193 193 193 > > > > >    cglen 74 74 79 > > > > >    fglen 63 64 60 > > > > >    fgcent mol 1 > > > > >    cgcent mol 1 > > > > >    mol 1 > > > > >    npbe > > > > >    bcfl mdh > > > > >    ion charge 1 conc 0.075 radius 2.0 > > > > >    ion charge -1 conc 0.075 radius 2.0 > > > > >    pdie 2.0 > > > > >    sdie 78.0 > > > > >    chgm spl4 > > > > >    srfm mol > > > > >    srad 1.4 # does not matter > > > > >    swin 0.3 # irrelevant with srfm mol > > > > >    sdens 10.0 # apbs suggested value > > > > >    temp 300.0 > > > > >    calcenergy total > > > > >    calcforce no > > > > >    write pot dx pot > > > > > end > > > > > > > > > > > > > > > quit > > > > > > > > > > > > > > > -- > > > > > Tu ne cede malis, sed contra audentior ito. -- Virgil > > > > > > > > > > --------------------------------------------------------------- > ---- > > > ----------- > > > > > This SF.net email is sponsored by Sprint > > > > > What will you do first with EVO, the first 4G phone? > > > > > Visit sprint.com/first -- http://p.sf.net/sfu/sprint-com-first > > > > > _______________________________________________ > > > > > apbs-users mailing list > > > > > apbs-users@... > > > > > https://lists.sourceforge.net/lists/listinfo/apbs-users > > > > > > > > > > > > > ----------------------------------------------------------------- > ---- > > > --------- > > > > This SF.net email is sponsored by Sprint > > > > What will you do first with EVO, the first 4G phone? > > > > Visit sprint.com/first -- http://p.sf.net/sfu/sprint-com-first > > > > _______________________________________________ > > > > apbs-users mailing list > > > > apbs-users@... > > > > https://lists.sourceforge.net/lists/listinfo/apbs-users > > > > > > > > > > > > -- > > > Tu ne cede malis, sed contra audentior ito. -- Virgil > > > > > > ------------------------------------------------------------------- > ---- > > > ------- > > > This SF.net email is sponsored by Sprint > > > What will you do first with EVO, the first 4G phone? > > > Visit sprint.com/first -- http://p.sf.net/sfu/sprint-com-first > > > _______________________________________________ > > > apbs-users mailing list > > > apbs-users@... > > > https://lists.sourceforge.net/lists/listinfo/apbs-users > > > > -- > Tu ne cede malis, sed contra audentior ito. -- Virgil ```