[gmx-users] atom types

Justin A. Lemkul jalemkul at vt.edu
Tue Aug 2 14:44:55 CEST 2011

Gavin Melaugh wrote:
> I hope this doesn't come across as stupid, or worse insolent. But what
> is the point in stating the charges in the atom type section then ?

They're probably a holdover from some earlier Gromacs version (an ancient one) 
that used those charges somehow, or they were listed there as some grand idea to 
have a more universal, streamlined topology builder that could use generic 
charges.  Just a guess.  They stayed there because it's far easier to just 
ignore the unnecessary information than the re-write the code and those files 
such that the lines are read properly with one less field.  The charges can be 
useful for starting parameterization; they are based on common functional groups 
that may not be far off from the correct values in many cases.


> Gavin
> Mark Abraham wrote:
>> On 02/08/11, *Gavin Melaugh * <gmelaugh01 at qub.ac.uk> wrote:
>>> Hi Justin
>>> I see that we may have got our wires crossed from the off.
>>> Consider the [pairs] directive, which determines which atoms interact
>>> in a 1,4 manner.  Consider two atoms listed the [pairs] directive. From
>>> the point of the Coulombic interaction between these two atoms I suppose
>>> my original question should have been: Does mdrun, when calculating the
>>> Coulombic potential between these two atoms, use the charges assigned to
>>> the atoms in the [atomtypes] directive or [atoms] directive ?
>> The charges from [atoms] are used. The charges in [atomtypes] are
>> never used in any forcefield currently used with GROMACS. It might not
>> be possible for grompp to ever use them, but I'd have to check the
>> code for that.
>> Mark
>>> Cheers
>>> Gavin
>>> Justin A. Lemkul wrote:
>>>> Gavin Melaugh wrote:
>>>>> Yes I think the example vindicates what I am saying as well. I
>>> suppose I
>>>>> the "contradiction" ( I'll call it the point of confusion) you
>>> refer to
>>>>> is perhaps when Justin (who is always more than helpful) said that
>>>>> "Charges are irrelevant for generation of pair interactions. 
>>> Nonbonded
>>>>> pair interactions are LJ, not Coulombic.  You will certainly have 1-4
>>>>> Coulombic interactions, but they are not generated by gen_pairs.  See
>>>>> manual section 5.3.4."
>>>> Charges *are* irrelevant - the information is not used when generating
>>>> pairs, which I thought was the original question.  The charge
>>>> information is used during MD, when the pair list tells mdrun which
>>>> atoms interact in what way. Then you get Coul. 1-4 terms.  Perhaps I
>>>> missed your point, but this whole thread started as "which charge is
>>>> used to generate pairs?"  The answer is still none.
>>>>> My sequence of 1-4 interaction generation should go like this I
>>> suppose:
>>>>> e.g.
>>>>> [pairs]
>>>>> 3      6
>>>>> no parameters present  therefore get from [pairtypes] directive.
>>>>> no [pairtypes] directive therefore get from [non_bonded parameters]
>>>>> directive as gen pairs = yes
>>>>> again no [non_bonded parameters] directive.
>>>>> Therefore generate 1,4 interaction parameters based on the normal
>>> sigma
>>>>> and epsilon values (comb rule 3) present in [atomtypes] directive, in
>>>>> accordance with fudge LJ and QQ.
>>>>> My point is, then in conclusion, that in this way surely the 1,4
>>>>> electrostatic interactions are determined by the pair list and in my
>>>>> case gen_pairs = yes.... no?
>>>> Yes.
>>>> -Justin
>>>>> Many Thanks
>>>>> Gavin
>>>>> Mark Abraham wrote:
>>>>>> On 02/08/11, *Gavin Melaugh * <gmelaugh01 at qub.ac.uk> wrote:
>>>>>>> Hi Mark
>>>>>>> Thanks for the reply.
>>>>>>> I am currently reading that section of the manual and, unless I am
>>>>>>> completely mistaken, it seems to vindicate what I am saying.
>>>>>>> "Extra Lennard-Jones and electrostatic interactions between pairs of
>>>>>>> atoms in a molecule can be added in the [pairs] section of a
>>> molecule
>>>>>>> definition".
>>>>>>> In my [atom types] directive I have atomtype, charge mass, sigma and
>>>>>>> epsilon etc. All nonbonding parameters are then calculated
>>>>>>> according to
>>>>>>> the combination rule (in my case 3). 1-4 interactions are then
>>>>>>> calculated based on the information in [pairs] directive (all atoms
>>>>>>> are
>>>>>>> three bond away). I just have the atom indices of each pair in this
>>>>>>> directive therefore with gen_pairs = yes, the interaction parameters
>>>>>>> between each pair (which are 1-4) are calculated based on Fudge
>>> LJ and
>>>>>>> Fudge QQ (which are both 0.5 in my case). All of this in conjunction
>>>>>>> with nrexcl =3.
>>>>>> That will generate parameters for the interactions listed in [pairs]
>>>>>> that do not have corresponding [pairtypes]. FudgeLJ and
>>>>>> [nonbond_params] are used in such generation, per other parts of 5.7.
>>>>>>>   Or am I completely wrong?
>>>>>>> In my set up then, are 1-4 Coulombic interactions determined by the
>>>>>>> pair
>>>>>>> list and fudge QQ?
>>>>>> If the contradiction you think exists is this one...
>>>>>>>> On 02/08/11, *Gavin Melaugh * <gmelaugh01 at qub.ac.uk> wrote:
>>>>>>>>> Hi Justin
>>>>>>>>> Again thanks for the reply. I am not disagreeing with you but If I
>>>>>>> don't
>>>>>>>>> include a [pairs] directive in the topology file (with gen_pairs
>>>>>>> =yes),
>>>>>>>>> then there are no 1-4 LJ nor 1-4 Coulombic energies written in
>>>>>>>>> the log
>>>>>>>>> file. When I include the [pair s] directive then both types of
>>>>>>>>> interaction are written to the log file. Therefore does gen_pairs=
>>>>>>> yes +
>>>>>>>>> [pairs] directive generate 1,4 LJ and 1,4 Coulomb according to
>>>>>>> fudge LJ
>>>>>>>>> and QQ?
>>>>>> ... then 5.3.4 indicates that the presence of a [pairs] directive
>>> will
>>>>>> generate the 1,4 output fields. The parameters for that output are
>>>>>> taken from [pairtypes]. If gen-pairs=yes then the parameters are
>>>>>> generated, else some warning/error occurs. The example in 5.7.1 has
>>>>>> some more explanation about the use of the fudge parameters.
>>>>>> Mark
>>> -- 
>>> gmx-users mailing list    gmx-users at gromacs.org
>>> http://lists.gromacs.org/mailman/listinfo/gmx-users
>>> Please search the archive at
>>> http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
>>> Please don't post (un)subscribe requests to the list. Use the
>>> www interface or send it to gmx-users-request at gromacs.org.
>>> Can't post? Read http://www.gromacs.org/Support/Mailing_Lists


Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080


More information about the gromacs.org_gmx-users mailing list