[gmx-users] Manual refinement of ATB topologies ?

[Sim gmx]

Hi all,

I would like to know what you consider to be the ideal refinement for a
topology created by ATB ?
How far should we go in the manual refinement ?

For instance, ATB gave me a topology for a molecule with an unsaturated
acyl chain. Here are some potential issues with this topology:

1) Three methyl groups at the end of the acyl chain (that are "3 methyl
groups away" from a double bond) carry a small charge (with a net group
charge of 0, however)
2) For the double bonds, parameters are "new parameters" (not found in
ffbonded.itp file)
3) Some bonded parameters in the ATB topology do exist in ffbonded.itp file
but are not supposed to be applied for this kind of bond (example: O - S
bond parameters are written for a C (involved in a double bond) - CH2 bond)

My "instinctive" behavior would be to:

1) Put a 0 charge to these methyl groups since they should not be charged
imo
2) Look for another gromos53a6 topology that includes a double bond (like
this one https://lipidbook.bioch.ox.ac.uk/package/show/id/16.html) and take
the double bond parameters from it
3) Replace those bonds by correct "default parameters" found in
ffbonded.itp file, in this case by a "C, CHn - C, CHn" bonded parameter.

However, I don't wish to break an "equilibrium" that ATB could have built
with some apparently weird parameters.

Hence, is it better to take the ATB topology as it comes, or to try to
refine it ? If the latter, how far should we go ? Isn't it risky or
complicated to justify (in a reviewing process, for example :)) ?

Thank you !