Easily lookup the origin of the opls_XXX atom types?

User discussions
,
1

GROMACS version: 2022.5-Debian_2022.5_2 , but I think applies to all versions
GROMACS modification: No

I have modified the x2top definition file for a much more complicated molecule.
(I’m placing this in the-postscript, as I also wrote some notes on how to do this, for others who are trying to model non-standard things.)

Our use case is a bit strange, in that we don’t actually care for the quality of the resulting force-field: we want a bonafide GROMACS topology as an input for the GRAPPA machine-learning model.

So all we really need to get correct are the non-bonded parameters, for which we need to choose the correct opls_XXX atom type.

Is there a look-up table somewhere for the GROMACS opls_XXX atom types, back to the original papers where they are defined?

Many years ago when I last did this, I had to cross-reference the actual Lennard-Jones parameters of an opls_XXX atom type, with the table in the original papers, to figure out which is which.
Is there a more structured way to look up the opls_XXX type?

P.S.
atomname2type.n2t

#Atom Type(.itp) Charge? Mass  Bonds    ((bonded atoms, distances))
C    opls_157    -0.18	12.011	4    H 0.108   H 0.108   H 0.108   C 0.150
C    opls_158    -0.12  12.011	4    H 0.108   H 0.108   C 0.150   C 0.150
C    opls_157    0.145  12.011	4    H 0.108   O 0.141   H 0.108   C 0.150
C    opls_157    0.145  12.011	4    H 0.108   H 0.108   O 0.108   C 0.150
C    opls_158    0.205  12.011	4    H 0.108   C 0.150   O 0.140   C 0.150
C    opls_158    -0.06  12.011	4    H 0.108   C 0.150   C 0.150   C 0.150   

C    opls_145    -0.12  12.011	3    C 0.150   C 0.150   H 0.108
C    opls_145    -0.12  12.011 	3    C 0.133   C 0.150   O 0.140
C    opls_235    0.5    12.011 	3    C 0.133   N 0.140   O 0.140
C    opls_235    0.5    12.011 	3    C 0.133   N 0.132   C 0.150
C    opls_235    0.5    12.011 	3    C 0.133   N 0.132   H 0.108

O    opls_236    -0.5   15.9994 1    C 0.123

N    opls_238    -0.5   14.0067 3    H 0.108   C 0.140   C 0.150
N    opls_238    -0.5   14.0067 3    H 0.108   C 0.132   N 0.123
N    opls_238    0      14.0067 2    C 0.140   N 0.123

H    opls_140    0.06    1.008 	1    C 0.108
H    opls_155    0.418   1.008	1    O 0.095
H    opls_241    0.30    1.008	1    N 0.095

O    opls_154    -0.683 15.9994 2    C 0.140   H 0.095
P    opls_393    1     30.97376 4    O 0.180   O 0.180   O 0.180   O 0.180
O    opls_395    -1     15.9994	2    P 0.180   H 0.095
O    opls_394    -1     15.9994 1    P 0.180
N    opls_237    0      14.0067 4    C 0.140   C 0.140   C 0.140   C 0.140

# x2top hacking
#
# x2top purely uses a distance critereon to define bonding.
#
# Key line is in mk_bonds,
#                 && (std::fabs(blen - nmt[i].blen[j]) <= 0.1
#                 * nmt[i].blen[j]))
#
# You see it is reading the t_nm2type nmt[] for the bond
# definition, and seeing if the bond length is within 10% of
# this distance.
#
# nmt[] is essentially this file, put into a C array
#
# So x2top will bond things it recognises from this file.
# Helpfully it tells you the element and atom number for things it doesn't
# recognise. 
#
# So if you have your molecule of interest open in an editor (such as Pymol, or
# probably more ideally Gaussview) that can tell you the atom index, and can
# calculate the distances for the nearby bonds, you can quickly write a line
# adding the local chemical space and defining what the opls_XXX type should
# be. 

# Frost group additions, for organic electron materials ==> Grappa
#
# NB: I just went with the first opls type I spotted to get this working!
# Requires more thought.

F   opls_719    0       18.998  1    C 0.14 

# aromatic carbons buried in the fused ring
C   opls_235    0.0    12.011  3    C 0.133 C 0.133 C 0.133
N   opls_238    0.0     14.0067 3    C 0.142   C 0.142 C 0.150  

# alphatic carbon dangling off the nitrogen
C   opls_158    -0.12  12.011	4    H 0.108   H 0.108   C 0.150   N 0.150

# thiophone sulphur
S   opls_202   -0.335   32.06   2    C 0.17 C 0.17
# BT sulphur; BT nitrogen
S   opls_202   -0.335   32.06   2    N 0.16 N 0.16
N   opls_251   -0.49   14.00670   2    S 0.16 C 0.13

# C-C#N terminating thingies - will these even exist as OPLS atom types?
N   opls_251   -0.49   14.00670   1      C 0.12
C   opls_235   0.0   12.0011   2      N 0.12 C 0.14

# carbons as part of the main flat fused rings with sulphur
C   opls_145    -0.12  12.011	3    C 0.14 C 0.14   S 0.17

# carbons, as in benzene, but with a flourine 
C   opls_145    -0.12  12.011	3    C 0.14 C 0.14   S 0.17
C   opls_145    -0.12  12.011	3    C 0.150   C 0.150   F 0.13
2

I’ve reminded myself that the most useful source of references for where the OPLS parameters have come from is in atomtypes.atp .

; OPLS atom types and masses.
; Atom types are named opls_X, where X is the OPLS number.
; The opls_ prefix is to avoid users confusing atom types
; (always prefixed) with atom numbers in molecules (never prefixed).
;
; Types 1-134 are from the united-atom OPLS, which can be
; useful for solvents and/or CH2 optimizations (e.g. in lipids).
; Explicit all-atom parameters start with opls_135.
; Note: For UA amide parameters -
; NMA - types 1,2,3,4,7,39
; Formamide 131,2,12,13
; DMF       131,2,3,132
; Acetamide 1,2,7,12,13
;
; Types 1-65 are united-atom parameters for proteins,
; see JACS 110, 1657 (1988).
;

But I’d still be very interested if there was a more automated manner of cross-referencing!