<div dir="ltr"><br><br><div class="gmail_quote"><div dir="ltr">On Sat, May 27, 2017 at 10:17 AM Igor Leontyev <<a href="mailto:ileontyev@ucdavis.edu">ileontyev@ucdavis.edu</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Hi,<br>
It is unlikely that inaccuracy in 4th digit of the parameters could<br>
introduce significant error in simulation results.<br>
<br>
However, for the energy comparison and debugging which is often made by<br>
curious people (including myself) when switching between Amber and<br>
Gromacs codes the inaccuracy is really bothering. Accepting suggested<br>
correction will match Amber and Gromacs code energies and make the<br>
comparison perfect (possibly for the cost of one time issue with<br>
regression test).<br></blockquote><div><br></div><div>True, but also makes GROMACS different from past versions of itself, which also would worry people doing such a comparison when e.g. contemplating upgrading their version. We're all for doing things correctly, but that should mean everybody actually implementing the published/accepted value, not agreeing with each other on how to be different.</div><div><br></div><div>It sounds like we're happy enough to change the values in GROMACS to match the paper - can people please raise the issue/question with others who distribute parameter sets?</div><div><br></div><div>Mark</div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Igor<br>
<br>
><br>
> Hi,<br>
><br>
> These difference are irrelevant for any practical purposes, since the<br>
> model (and sampling) is not by far that accurate. But it would still be<br>
> nice to have all decimals that we have in files correct.<br>
> Is this somehow related to the differences in the kcal/mol to kJ/mol<br>
> conversion in different packages and/or at different times?<br>
><br>
> Cheers,<br>
><br>
> Berk<br>
><br>
> On 24/05/17 21:18 , Igor Leontyev wrote:<br>
>> Hi all. The issue was discussed in past but nothing was corrected<br>
>> since then.<br>
>><br>
>> <a href="https://mailman-1.sys.kth.se/pipermail/gromacs.org_gmx-developers/2011-April/005178.html" rel="noreferrer" target="_blank">https://mailman-1.sys.kth.se/pipermail/gromacs.org_gmx-developers/2011-April/005178.html</a><br>
>><br>
>><br>
>> To make gromacs TIP3P energies matching Amber code energies:<br>
>>> Suggested correction in amberXX.ff/ffnonbonded.itp:<br>
>>> ;OW 8 16.00 0.0000 A 3.15061e-01 6.36386e-01<br>
>>> OW 8 16.00 0.0000 A 3.15076e-01 6.35968e-01<br>
>><br>
>> Igor<br>
>><br>
>>> So, going back to the original TIP3P paper, the parameters are listed<br>
>> as A and C parameters (in the formula a/r^12 ? b/r^6).<br>
>>><br>
>>> The parameters listed there are A = 582000 A^12 kcal/mol and C = 595<br>
>>> A^6 kcal/mol.<br>
>>><br>
>>> Converting to sigma/epsilon form, this gives:<br>
>>> Sigma = 0.31506561105 nm<br>
>>> Epsilon = 0.6362717354 kJ/mol<br>
>>><br>
>>> This compares with the values in the most of the distributed ITP<br>
>>> files of<br>
>>><br>
>>> Sigma = 0.315061 nm (a difference of 4.6 x 10^-6 nm, or a relative<br>
>>> difference of of 1.4 x 10^-5)<br>
>>> Epsilon = 0.636386 kJ/mol (a difference of 1.1 x 10^4 kJ/mol, or a<br>
>>> relative difference of 1.8 x 10^-4)<br>
>>> (note the GROMACS one is also 0.152100 kcal/mol)<br>
>>> I don?t know how much of a difference this would make (likely a<br>
>>> difference of the scale of).<br>
>>><br>
>>> Now, that?s in the original paper. What values are ACTUALLY<br>
>>> specified by other codes?<br>
>>><br>
>>> For AMBER, (as produced by tleap), it outputs.<br>
>>><br>
>>> A = 581935.564 A^12 kcal/mol<br>
>>> C = 594.825035 A^6 kcal/mol<br>
>>><br>
>>> Which leads to:<br>
>>><br>
>>> Sigma = 0.315075 (a difference of 3.0 x 10^-5 nm, or a relative<br>
>>> difference of 3.0 x 10^-5)<br>
>>> Epsilon = 0.635968 kJ/mol (a difference of 9.4 x 10^4 kJ/mol or a<br>
>>> relative difference of 4.7 x 10^-4)<br>
>>> (which is also 0.152000 kcal/mol)<br>
>>><br>
>>> So ? I don?t know. Nobody seems to be getting it _exactly_ right,<br>
>>> and it?s also not so clear how much of a difference it makes. One<br>
>>> can estimate that enthalpies would have a relative error of around<br>
>>> 10^-4, and relative errors in the density of 3*dsigma = 10^-4 to 10^-5.<br>
>>><br>
>>> So I don?t know that there is anything actionable to take. Just<br>
>>> thought it was interesting to bring out. Something to ponder for now.<br>
>><br>
<br>
><br>
> I?d say ?probably? irrelevant for practical purposes. It wouldn?t take too much simulation time to determine the density sensitively enough to tell the difference. But, pretty much irrelevant.<br>
><br>
> ? Is this somehow related to the differences in the kcal/mol to kJ/mol<br>
> ? conversion in different packages and/or at different times?<br>
><br>
> I think that it?s due to differences in rounding from the original A and C values vs. rounding from conversion of them to epsilon and sigma values. The epsilons in AMBER and GROMACS are clearly equal to using either 0.152100 or 0.15200 kcal/mol as the value (the parameters in the original paper correspond to 0.151207).<br>
><br>
> I think there might have been a kJ/kcal issue at some point, but those were fixed a decade or so ago (I seem to recall that popping up when I was a graduate student, but not since then).<br>
><br>
> ? These difference are irrelevant for any practical purposes, since the<br>
> ? model (and sampling) is not by far that accurate. But it would still be<br>
> ? nice to have all decimals that we have in files correct.<br>
><br>
> At this stage, it probably involves getting some general set everyone together and deciding in what the ?right? values are. I could potentially do that . . . but as Berk points out, not that big a deal. Unless one is trying to match energy snapshots, and getting frustrated.<br>
><br>
> For what it?s worth, the Desmond viparr utility uses for TIP3P: {"sigma": 3.15061, "epsilon": 0.1521}. Both match GROMACS .itp libraries, but not tleap or the original paper.<br>
><br>
> Best,<br>
> ~~~~~~~~~~~~~~~~<br>
> Michael Shirts<br>
> Associate Professor<br>
> <a href="mailto:michael.shirts@colorado.edu" target="_blank">michael.shirts@colorado.edu</a><br>
> <a href="http://www.colorado.edu/lab/shirtsgroup/" rel="noreferrer" target="_blank">http://www.colorado.edu/lab/shirtsgroup/</a><br>
> Phone: (303) 735-7860<br>
> Office: JSCBB C123<br>
> Department of Chemical and Biological Engineering<br>
> University of Colorado Boulder<br>
><br>
> On 5/25/17, 1:51 PM, "<a href="mailto:gromacs.org_gmx-developers-bounces@maillist.sys.kth.se" target="_blank">gromacs.org_gmx-developers-bounces@maillist.sys.kth.se</a> on behalf of Berk Hess" <<a href="mailto:gromacs.org_gmx-developers-bounces@maillist.sys.kth.se" target="_blank">gromacs.org_gmx-developers-bounces@maillist.sys.kth.se</a> on behalf of <a href="mailto:hess@kth.se" target="_blank">hess@kth.se</a>> wrote:<br>
><br>
> Hi,<br>
><br>
> These difference are irrelevant for any practical purposes, since the<br>
> model (and sampling) is not by far that accurate. But it would still be<br>
> nice to have all decimals that we have in files correct.<br>
> Is this somehow related to the differences in the kcal/mol to kJ/mol<br>
> conversion in different packages and/or at different times?<br>
><br>
> Cheers,<br>
><br>
> Berk<br>
><br>
> On 24/05/17 21:18 , Igor Leontyev wrote:<br>
> > Hi all. The issue was discussed in past but nothing was corrected<br>
> > since then.<br>
> ><br>
> > <a href="https://mailman-1.sys.kth.se/pipermail/gromacs.org_gmx-developers/2011-April/005178.html" rel="noreferrer" target="_blank">https://mailman-1.sys.kth.se/pipermail/gromacs.org_gmx-developers/2011-April/005178.html</a><br>
> ><br>
> ><br>
> > To make gromacs TIP3P energies matching Amber code energies:<br>
> > > Suggested correction in amberXX.ff/ffnonbonded.itp:<br>
> > > ;OW 8 16.00 0.0000 A 3.15061e-01 6.36386e-01<br>
> > > OW 8 16.00 0.0000 A 3.15076e-01 6.35968e-01<br>
> ><br>
> > Igor<br>
> ><br>
> >> So, going back to the original TIP3P paper, the parameters are listed<br>
> > as A and C parameters (in the formula a/r^12 ? b/r^6).<br>
> >><br>
> >> The parameters listed there are A = 582000 A^12 kcal/mol and C = 595<br>
> >> A^6 kcal/mol.<br>
> >><br>
> >> Converting to sigma/epsilon form, this gives:<br>
> >> Sigma = 0.31506561105 nm<br>
> >> Epsilon = 0.6362717354 kJ/mol<br>
> >><br>
> >> This compares with the values in the most of the distributed ITP<br>
> >> files of<br>
> >><br>
> >> Sigma = 0.315061 nm (a difference of 4.6 x 10^-6 nm, or a relative<br>
> >> difference of of 1.4 x 10^-5)<br>
> >> Epsilon = 0.636386 kJ/mol (a difference of 1.1 x 10^4 kJ/mol, or a<br>
> >> relative difference of 1.8 x 10^-4)<br>
> >> (note the GROMACS one is also 0.152100 kcal/mol)<br>
> >> I don?t know how much of a difference this would make (likely a<br>
> >> difference of the scale of).<br>
> >><br>
> >> Now, that?s in the original paper. What values are ACTUALLY<br>
> >> specified by other codes?<br>
> >><br>
> >> For AMBER, (as produced by tleap), it outputs.<br>
> >><br>
> >> A = 581935.564 A^12 kcal/mol<br>
> >> C = 594.825035 A^6 kcal/mol<br>
> >><br>
> >> Which leads to:<br>
> >><br>
> >> Sigma = 0.315075 (a difference of 3.0 x 10^-5 nm, or a relative<br>
> >> difference of 3.0 x 10^-5)<br>
> >> Epsilon = 0.635968 kJ/mol (a difference of 9.4 x 10^4 kJ/mol or a<br>
> >> relative difference of 4.7 x 10^-4)<br>
> >> (which is also 0.152000 kcal/mol)<br>
> >><br>
> >> So ? I don?t know. Nobody seems to be getting it _exactly_ right,<br>
> >> and it?s also not so clear how much of a difference it makes. One<br>
> >> can estimate that enthalpies would have a relative error of around<br>
> >> 10^-4, and relative errors in the density of 3*dsigma = 10^-4 to 10^-5.<br>
> >><br>
> >> So I don?t know that there is anything actionable to take. Just<br>
> >> thought it was interesting to bring out. Something to ponder for now.<br>
--<br>
Gromacs Developers mailing list<br>
<br>
* Please search the archive at <a href="http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List" rel="noreferrer" target="_blank">http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List</a> before posting!<br>
<br>
* Can't post? Read <a href="http://www.gromacs.org/Support/Mailing_Lists" rel="noreferrer" target="_blank">http://www.gromacs.org/Support/Mailing_Lists</a><br>
<br>
* For (un)subscribe requests visit<br>
<a href="https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developers" rel="noreferrer" target="_blank">https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developers</a> or send a mail to <a href="mailto:gmx-developers-request@gromacs.org" target="_blank">gmx-developers-request@gromacs.org</a>.<br>
</blockquote></div></div>