<html>
<head>
<style>
.hmmessage P
{
margin:0px;
padding:0px
}
body.hmmessage
{
font-size: 10pt;
font-family:Verdana
}
</style>
</head>
<body class='hmmessage'>
Hi,<br><br>I guess that what you want to do is exactly what the decouple mdp options do.<br>You probably don't need modify any topology files.<br>If couple-intramol=no (default) than all intra-molecular interactions are not turned<br>off and are plain LJ/Coulomb without cut-off.<br><br>Berk<br><br>> Date: Wed, 28 Jan 2009 00:07:43 -0500<br>> From: chris.neale@utoronto.ca<br>> To: gmx-users@gromacs.org<br>> Subject: [gmx-users] Re: decoupling charge while maintaining intramolecular potentials<br>> <br>> Thank you Michael for the detailed reply, I have posted clarifications <br>> and additional questions or remarks inline below.<br>> <br>> > Hi, Chris-<br>> ><br>> > I unfortunately can't be too much help, because my free energy<br>> > calculations are done through a modified version of Gromacs 3.1.4, and<br>> > I am currently working with Berk and Erik to get the important<br>> > modifications into the 4.0 branch.<br>> ><br>> >> I am a new user of the free energy code. I am somewhat confused<br>> >> regarding the method that should be applied to decouple the long range<br>> >> interactions of the solvent from the solute while still maintaining<br>> >> intramolecular long-range interactions for the solute.<br>> ><br>> > By long range interactions, you mean the ones outside the cutoff,<br>> > correct? Because there's another set of difficulties dealing with the<br>> > ones that are shortange nonbonded interactions as well. Could you<br>> > clarify?<br>> <br>> When I said "long-range" I should have said "non-bonded" instead <br>> throughout this entire document. Sorry for the incorrect terminology. <br>> Therefore I am under the impression that one wants to decouple the <br>> intermolecular solvent-solute nonbonded interactions without affecting <br>> any intramolecular nonbonded interactions. I have seen this applied in <br>> your papers and some others, many of them done via charmm.<br>> <br>> ><br>> >> I was able to find some information in this thread:<br>> >> http://www.gromacs.org/pipermail/gmx-developers/2006-January/001498.html<br>> >> but it is still unclear to me.<br>> ><br>> > I'm not surprised, it was unclear for us at the time as well :)<br>> ><br>> >> I have reproduced the methane and tip3p energies of solvation based on<br>> >> the tutorial that is on the gromacs wiki. In this case, I simply<br>> >> assigned new charge values of 0 in the B state without making any<br>> >> special considerations for intramolecular O-H values. However, tip3p<br>> >> is rigid and perhaps maintaining the intramolecular q-q and LJ<br>> >> components is not essential in this case.<br>> ><br>> > This is where I get confused -- are you talking about long range, or<br>> > any intramolecule interactions. Certainly for methane and tip3p,<br>> > there aren't any, because all atoms are 1,3 neighbors.<br>> <br>> I am talking about any nonbonded intramolecular interactions. Further, <br>> your point regarding 1-2 and 1-3 interactions is well taken. My <br>> molecule of interest is a dodecylphosphocholine detergent and <br>> therefore there will be substantial intramolecular nonbonded <br>> interactions that I believe it would be best to maintain.<br>> <br>> ><br>> >> 1. Is it necessary to maintain the intramolecular long-range<br>> >> interactions for the solute while decoupling LJ or charge? If not<br>> >> absolutely required, does it affect the rate of convergence?<br>> ><br>> > If you change the intramolecular nonbonded interactions, then you<br>> > would have to perform a second vacuum calculation in which you turn<br>> > them back on. In terms of rates of convergence -- nobody really<br>> > knows. If it's a intramolecular hydrogen bonding system, then turning<br>> > off the intramolecular interactions might be faster.<br>> <br>> What I am actually doing is annihilating one DPC detergent from a DPC <br>> micelle and, separately, annihilating one DPC detergent monomer from <br>> bulk water. The dG(monomer) minus dG(micelle) should then be related <br>> to the relative probabilities of monomeric and aggregated states. I <br>> have left out the monomer restraints and the volume correction here <br>> for simpicity. It is my impression that this monomer-in-bulk <br>> annihilation provides the second simulation that you mentioned above, <br>> although I will still require this second simulation even if I <br>> maintain my intramolecular non-bonded interactions since I am <br>> interested in the relative probabilities of the monomeric and <br>> aggregated state, both in solution, and not in solvation free energies.<br>> <br>> The reason that I thought one would want to not change the <br>> intramolecular interactions is that it seems to me that this will make <br>> the d(pot)d(lambda) energies larger and then subtracting 200-190=10 <br>> seems like it will have more uncertainty than if the dvdl integrated <br>> energies were all smaller. This is just based on the idea that a small <br>> difference of two large numbers is often difficult to obtain precisely.<br>> <br>> However, I see your point about relatively stable intramolecular <br>> interactions. I will need to think further about that one.<br>> <br>> ><br>> >> 2. Is this already handled by the free-energy code?<br>> ><br>> > I can't speak for the 4.0 code. Berk was introducing nonbonded pair<br>> > terms such that these pair terms would overrule the 'alchemical'<br>> > transformation, resulting in unchanged intramolelcular nonbonded<br>> > interactions. I actually don't know the current state of this change,<br>> > though.<br>> ><br>> >> 3. If not, how might one go about doing this? My confusion with some<br>> >> additional [ pairs ] entry is how gromacs would get the right<br>> >> combination for lambda=0 and lambda=1 (not to mention intermediate<br>> >> states).<br>> ><br>> > I'm a bit confused. I would think that you would just want to set<br>> > them to the lambda=0 state, so that intramolecular interactions were<br>> > preserved. Am I thinking of something different than you are?<br>> <br>> If you're just talking about how to handle the lambda=1 simulation by <br>> setting pairs according to the nonbonded values that are achieved in <br>> the lambda=0 state, then I believe that we are thinking about the same <br>> thing. However, it gets messy for intermediate values of lambda (not <br>> to mention impossible for the charging simulations) so I'll outline my <br>> logic more elaborately below.<br>> <br>> Assume that the 4.0.3 free-energy code uses lambda to scale all <br>> nonbonded interactions including intreamolecular ones for the <br>> annihilated molecule. Further assume that I create a new [ pairs ] <br>> section that entirely accounts for all intramolecular LJ interactions <br>> of the DPC to be annihilated (let's talk LJ annihilation only to make <br>> this part simpler). If I include this new pairs section for lambda=1, <br>> then I expect no solute-solvent nonbonded interactions but to maintain <br>> regular intramolecular nonbonded interactions from the pairs section. <br>> However, if I include this new pairs section for lambda=0.5, then I <br>> expect to have 1.5x strength nonbonded intramolecular interactions for <br>> the annihilated solute (once from pairs and one-half from regular <br>> nonbonded). This would require different .itp files for each lambda <br>> value with pairs set according to:<br>> <br>> pair(lambda)=pair(Astate)*(1-lambda) + pair(Bstate)*(lambda)<br>> <br>> if the conversion from A state to B state is linear and thus the new <br>> pairs section would be different for each lambda and would compensate <br>> for the loss of intramolecular nonbonded interactions as lambda is <br>> increased.<br>> <br>> Still, I believe that it is not yet possible to define coulombic <br>> interactions in the pairs section so this would not work for the <br>> coulombic portion.<br>> <br>> Thanks you,<br>> I sincerely appreciate your time in assisting me,<br>> Chris.<br>> <br>> _______________________________________________<br>> gmx-users mailing list gmx-users@gromacs.org<br>> http://www.gromacs.org/mailman/listinfo/gmx-users<br>> Please search the archive at http://www.gromacs.org/search before posting!<br>> Please don't post (un)subscribe requests to the list. Use the <br>> www interface or send it to gmx-users-request@gromacs.org.<br>> Can't post? Read http://www.gromacs.org/mailing_lists/users.php<br><br /><hr />What can you do with the new Windows Live? <a href='http://www.microsoft.com/windows/windowslive/default.aspx' target='_new'>Find out</a></body>
</html>