[gmx-users] clashes after energy minimization - Ser2 multiple copies

[David de Sancho]

Hi all
I have been having trouble setting up a peptide+water simulation system due
to the appearance of clashes in an energy minimization.

Trying to pin down the origin of the problem I have found that a minimum
system that reproduces my problem is a box with multiple copies of an
unblocked serine dipeptide, which I am simulating using a combination of an
Amber force field and TIP3P water. I have created a gist in case anyone
could please help.

https://gist.github.com/daviddesancho/00b23264fed0b6d20ec4e26e7d7810e7

At some point in the minimization, the energy drops asymptotically (also,
the forces become massive, ~1e+07). Surprisingly, this does not occur when
the number of copies of the dipeptide is N=1, 2 or 3. The resulting
minimized structure has a cycle formed by atoms in the C-terminal serine,
and one of the carboxylic oxygens overlaps with the alcoholic hydrogen.

So far, I have been able to reproduce this with different force fields from
the Amber family and the drop in the energy seems easy to rationalize,
based on the parameters. The sidechain hydrogen (HG) and oxygen (OC1) have
opposite charges, while the hydrogen has 0 sigma and epsilon. Hence the
interaction energy of the hydrogen will be zero and the Coulomb term will
be maximized for r_OH=0. No surprise, then that they attract each other.
But why do they clash only when there are 4 molecules around?

I have checked different box sizes and initial configurations to no avail.
Also, adding water and ions to neutralize the box and increase charge
screening does not change the outcome qualitatively (although it makes the
visualization trickier).

As always, all help greatly appreciated,


David