[gmx-users] Conversion of charmm36 parameters from namd to gromacs

[Justin Lemkul]

On 12/4/17 7:47 AM, Daniel Bauer wrote:
> Hello Justin,
>
> Thanks for your response. I was under the assumption that entries in [
> pairtypes ] are the ones that override standard LJ interactions derived
> by pairgen. I now know this is a different section.
>
>  From what you are saying I come to the conclusion that the POT O
> interaction in charmm36 for gromacs have no custom potential but are
> derived via combination rules (as they are in namd). Nevertheless, the
> numbers still dont seem to match:
>
> Since we can ignore 1-4 Interactions in charmm36 as well as gromacs for
> POT-O, the resulting parameters from the namd version are:
>
> sig = 0.3463750 nm; eps=0.42750571533 kJ/mol

Can you please share the source of these parameters? There should be no 
"NAMD version" of anything. The force field files that you should be 
using are just the CHARMM distribution. The GROMACS port of CHARMM36 
comes directly from those files. There is no official NBFIX for POT-O 
(yet, we do have one but it's not in the official distribution) so the 
combination rules should prevail.

-Justin

> However, combining the entries for O and POT in the atomtypes section of
> the gromacs version i get something different:
>
> [ atomtypes ]
> O          8    15.999000    0.000  A  0.302905564168  0.50208
> POT    19    39.098300    0.000  A  0.314264522824  0.36401
> =======================================
>
> sig = 0.308585043 nm; eps=0.42750689 kJ/mol
>
> Again, the epsilon value is matching the namd version while sigma is
> still off by about 0.4 A.
>
> If my conversion of the namd parameters and combination rules are
> correct, this would be a problem. The different position of the energy
> minimum results in a potential energy difference of around 7 kJ/mol for
> a distance of 0.29 nm (typical distance for POT-O in the selectivity
> filter of potassium channels, see attached image).
> Can you give me a source where I can read up on why the values for sigma
> and epsilon in gromacs are the way they are? It must be more than a
> direct conversion of the numbers from kcal to kJ. Or did I again made a
> mistake or took the wrong numbers?
>
> Best regards,
> Daniel
>
>
>
> On 12/03/2017 11:23 PM, Justin Lemkul wrote:
>>
>> On 12/3/17 4:18 PM, Daniel Bauer wrote:
>>> Hello,
>>>
>>>
>>> I compared the LJ parameters for the interaction Potassium - backbone
>>> carbonyl oxygen for CHARMM36 between the Gromacs and NAMD version of the
>>> forcefield. I found different numbers for the sigma value i cannot
>>> explain to myself:
>>>
>>>
>>> The stock charmm36 implementation has the following LJ parameters for
>>> potassium and oxygen:
>>>
>>> ; atom    Emin (kcal/mol)    Rmin/2 (A)
>>>
>>> POT    0.087    1.76375
>>>
>>> O    0.120    1.7
>>>
>>>
>>> Converting the above values to units used in gromacs (kJ/mol and nm) and
>>> applying standard combination rules this should give the following
>>> nonbonded energy parameters:
>>>
>>> [ pairtypes]
>>>
>>> ; sig = (Rmin,i/2 + Rmin,j/2)/10
>>>
>>> ; eps = sqrt(Emin,i * Emin,j) * 5.184 kcal/kJ
>>>
>>> ; i    j    func    sig (nm)    eps (kj/mol)
>>>
>>> O    POT    1    0.346 0.4275
>>>
>>> However, the actual entry for the interaction in GROMACS/charmm36 has
>>> the energy minimum at a much smaller distance (epsilon is on the point
>>> though!)
>>>
>>> ; i    j         func     sig (nm)        eps (kJ/mol)
>>>
>>> O    POT    1    0.282    0.4275
>>>
>>>
>>> Can somebody tell me the reason for this huge difference? Is there an
>>> error in my calculation? I am trying to convert an NBFIX applied to this
>>> specific interaction to gromacs. However, I am not sure how to proceed
>>> without doing the reparametrization in gromacs again, because the stock
>>> values differ that much.
>> This is actually a value that isn't used. [pairtypes] are for 1-4
>> interactions, which can never occur between a K+ ion and a carbonyl O.
>> The inclusion in the list is just because it's really hard to tell our
>> conversion program to exclude some massive list of impossible
>> interactions. So they get generated.
>>
>> But to your point, you're missing a critical point of the CHARMM
>> parameter format. The full NONBonded parameter entry in
>> par_all36m_prot.prm for carbonyl O is:
>>
>> O      0.000000  -0.120000     1.700000   0.000000  -0.120000 1.400000
>>
>> The last fields are epsilon and Rmin/2 when that atom type is used in
>> a 1-4 interaction. So -0.12/1.7 is for the normal nonbonded
>> interaction, -0.12/1.4 is for any 1-4 involving O.
>>
>> -Justin
>>
>
>

-- 
==================================================

Justin A. Lemkul, Ph.D.
Assistant Professor
Virginia Tech Department of Biochemistry

303 Engel Hall
340 West Campus Dr.
Blacksburg, VA 24061

jalemkul at vt.edu | (540) 231-3129
http://www.biochem.vt.edu/people/faculty/JustinLemkul.html

==================================================