Torsions around non-single C~S bonds

[bannanc]

Here are some example cases:

Aromatic bonds, given a weird case.

This molecule appears in the tripos mol2 files, but not in the parm@frosst typed drugbank molecules (which includes drugbank molecules assigned the generic atom type 'gg'). This is odd, but outside the point. Other than I don't know which parm@Frosst atomtypes it gets.

Here's the section for hypervalent sulfur torsions as it is now in smirffishFrcmod

# hypervalent S-C
[*:1]~[#16X4,#16X3+0:2]-[#6X4:3]-[*:4]  1  0.000    0.0     3  Frosst X_-SO-CT-X  guess
[#6X4:1]-[#16X4,#16X3+0:2]-[#6X4:3]-[#1:4]  1  0.250    0.0     1  Frosst CT-SO-CT-H1  RHF/6-41G(d,p) Dec 12 1997
[#6X4:1]-[#16X4,#16X3+0:2]-[#6X4:3]~[#6X4:4]  1  0.800    0.0     3  Frosst CT-SO-CT-CT  RHF/6-41G(d,p) Dec 12 1997
[*:1]~[#16X4,#16X3+0:2]-[#6X3:3]~[*:4]  1  1.750  180.0     2  Frosst X_-SO-CA-X  generic guess
[#6:1]-[#16X4,#16X3+0:2]-[#6X3:3]~[*:4]  1  0.750    0.0     1   Frosst generic

Note all of these have single bonds

These are the generic C-S bonds in GAFF2 that are labeled as double bonds:

X -c2-s4-X    4   26.600       180.000           2.000      c2=s4 double bond  !!!
X -ce-s4-X    4   26.600       180.000           2.000      c2=s4 double bond  !!!
X -cf-s4-X    4   26.600       180.000           2.000      c2=s4 double bond  !!!
X -c2-s6-X    6   39.900       180.000           2.000      c2=s6 double bond  !!!
X -ce-s6-X    6   39.900       180.000           2.000      c2=s6 double bond  !!!
X -cf-s6-X    6   39.900       180.000           2.000      c2=s6 double bond  !!!

The aromatic molecule doesn't make it to the GAFF2 stage of our project so I assume GAFF2 doesn't handle the aromatic sulfur? Or something else went wrong with the molecule.
I suggest we just put a generic double or aromatic bond in the top of the C-S torsion section. It will at least be better than the aromatic bond getting the generic.

I recommend we add this line:

[*:1]~[#16:2]=,:[#6:3]~[*:4]  1   6.5    180.0    2.0  c=s double bond from GAFF2 (CCB 4/4/2017)

We could make it even more generic with an any bond in the middle, but that doesn't seem good to me as any single bond not covered by the other section would be assigned this parameter.

[bannanc]

Added in pull request #43