[gmx-users] weird problem on small molecules with triple bonds

[David Mobley]

All,

I may be in a hole of my own digging, here, but I've been trying to
simulate a large set of small molecules in explicit solvent. I'm using
the Amber GAFF parameters, converted over to GROMACS using our
amb2gmx.pl script.

The problem I'm running into is that for anything with a triple bond
(i.e. propyne), I get "blowing up", even in vacuum (i.e., I get a
series of constraints warnings finally ending in a crash). This is
with a 2 fs timestep. Since of course things with triple bonds have
larger spring constants, I have tried reducing the timestep, but still
run into problems with timesteps as small as even 0.5 fs, which is
ridiculous, especially since the spring constants are not THAT much
bigger.

This made me suspect some sort of problem with the conversion script
with respect to triple bonds, so I went through and hand-checked all
of the bond, angle, and torsion parameters. Everything appears fine,
in that manual conversion from the GAFF values gives me the same
values the script is giving me.

I should mention that, if I run these simulations in AMBER, even with
2 fs timesteps, I don't have any problems.

Anyone have any thoughts? I'm a bit at a loss about what else to look at, here.

One other thing I notice -- which, again, doesn't appear to cause any
problems in AMBER --  is that GAFF specifies a dihedral force constant
of zero for the c1-c1-c3-H dihedral in propyne, which means that the
methyl can rotate rapidly. This doesn't appear to cause any problems
in AMBER, but I do notice that the LINCS warnings I get relate to
constraint deviations between c3 and H.

Any ideas will be appreciated.

Thanks,
David