unsafe atom types

[nhenriksen]

Some AMBER tools use atom types to make parameter/topology decisions. This obviously creates a roadblock if we want to treat atom types as arbitrarily. It appears the biggest problem is LEaP. For example, if uniq_types gives us a mol2 which lists a carbon as having atom type "h1", then LEaP will assume the atom is a hydrogen (for some reason it doesn't check the mass in the frcmod). This creates problems when you try to put four bonds to the carbon, because LEaP assumes that is a problem.

The solution appears to be: 1) loadamberparams, 2) loadmol2, 3) set Unit.Res.AtomName element ElementSymbol, 4) loadpdb with CONECT records

For example, methanol would look like:

loadamberparams mth.frcmod
MOL = loadmol2 mth.mol2
set MOL.MOL.C1 element C
set MOL.MOL.O1 element O
set MOL.MOL.H1 element H
set MOL.MOL.H11 element H
set MOL.MOL.H12 element H
set MOL.MOL.HC11 element H
MOL = loadpdb mth.pdb

I don't think parmed can write pdb with explicit CONECT records but hopefully pytraj can ... because cpptraj can.

[hainm]

hopefully pytraj can ... because cpptraj can.

indeed

pytraj.write_traj('hey.pdb', traj, options='conect model')