Re: dihedral parameter conversion

From: Markus Dahlgren (markus.dahlgren_at_yale.edu)
Date: Thu May 02 2013 - 10:08:51 CDT

Yes, the fourier coefficients retain their signs. from the OPLS file for
proteins, formatted in the charmm format, dihedrals are given in the
following way:

C271 C274 C136 C293 -1.5925 1 0.0
C271 C274 C136 C293 -0.4125 2 180.0
C271 C274 C136 C293 0.2465 3 0.0

The above term is the CT-CT-CT-C dihedral of protein sidechains (Lysine
for instance). I assume the last column is for a phase shift. The 6th
column indicates whether it is a V1, V2, V3, or V4 term. The signs have
to be preserved because if they would removed, the parameters would be
useless. In this case there are V1, V2, and V3 terms. V2 and V4 always
have a 180.0 in column 6 and V1 and V3 has a 0.0.

The dihedral coefficients (column 5) have been divided by 2 from the
standard OPLS format. I dont know how this is handled internally in
NAMD, but if you look at the OPLS file for Gromacs, the negative signs
are kept for the OPLS coefficients. Please let me know what you find :)

-Markus

On 5/2/2013 10:41 AM, JC Gumbart wrote:
> Do you use the negative coefficients in your charmm-formatted parameter file? There must be a reason though that the charmm force field has NO dihedral terms with negative coefficients. It seems like the two approaches are incompatible.
>
> I'll take a look at Dan Price's work to see if I can further sort it out.
>
> Thanks!
>
> -----Original Message-----
> From: owner-namd-l_at_ks.uiuc.edu [mailto:owner-namd-l_at_ks.uiuc.edu] On Behalf Of Markus Dahlgren
> Sent: Thursday, May 02, 2013 10:17 AM
> To: namd-l_at_ks.uiuc.edu
> Subject: Re: namd-l: dihedral parameter conversion
>
> There is an OPLS parameter file that Dan Price prepared using the program PEPZ that is output in the CHARMM format. The fourier coefficients are the OPLS coefficients divided by 2. So a V1 of -5 would be -2.5.
>
> You also have to convert sigma and epsilon for all atom types.
>
> I have run several NAMD simulations using OPLS and I have checked dihedral distributions generated with NAMD and compared it with MP2 dihedral profiles that parameters were fitted against and it looks correct. OPLS and CHARMM have very similar formats. The main difference is the absence of CMAP in OPLS.
>
> -Markus
>
> On 5/2/2013 9:17 AM, JC Gumbart wrote:
>> I'm not so familiar with the formats of force fields other than CHARMM, but I want to convert one for OPLS to CHARMM-style for running in NAMD. The main issue I've yet to resolve though is the format of the dihedrals. Here's an example line from the original parameter file:
>>
>> ; ai aj ak al funct ; Amber type OPLS type Type V1 V2 V3 Comments
>> 1 4 5 6 3 -0.50208 -1.50624 0.00000 2.00832 0.000 0.000 ; C3-N3-C2-C2 4031-4030-4032-4004 5000 0.000 0.000 -0.240
>>
>> I realize the first part is an RB format. The second part, I guess (please correct me if I'm wrong!!!), uses this functional form:
>>
>> V(φ) =V1(1+ cosφ)/2+V2(1−cos2φ)/2+V3(1+ cos3φ)/2+V4(1−cos4φ)/2
>>
>> So for the example line, the potential would be V = -0.12*(1+cos(3*phi)). But how to represent this in CHARMM??? Because it can't be JUST a phase shift, then we would have V = 0.12*(1+cos(3*phi-180)) = 0.12*(1-cos(3*phi)). In other words, there is a constant shift in the potential energy equal to V3.
>>
>> What simple fact am I misunderstanding here? How does one convert force constants less than zero to charmm, where they are always greater than zero?
>>
>> Thanks!
>> JC
>>

This archive was generated by hypermail 2.1.6 : Wed Dec 31 2014 - 23:21:10 CST