[gmx-users] Unnatural amino acid problems

[Mark Abraham]

> Hello
> I am parametrizing an unnatural amino acid, phenylglycine (PHG), in
> the OPLS FF in GROMACS. I have put in the necessary parameters in
> the .hdb, .rtp, and the aminoacids.dat file.
> 1. When I do grompp for the EM, I get an error saying that I have not
> put in the appropriate parameters for the bonds, angles and
> dihedrals.

Well that means you haven't done it :-) Go back and compare what you've
done for PHG with other amino acids, see what is different and fix it.

> Surprisingly, when I do put in the parameters (from the
> ffoplsaabon.itp file), the angles, bonds, and dihedrals get even more
> distorted, a counterintuitive observation. In fact the simulation is
> much better off when I ignore the warnings and the program puts in
> zeros after the required parameters in the .top file. Why is this
> happening?

You can't tell that your force field is a reasonable physical model by
looking at the results on their own - you need a comparison with
experiment. Often you can tell by looking at the results that it is
completely wrong, however. Just putting in zeroes guarantees you a
spurious model, whether you can recognise that from the results or not.

> Another question related to this:
> 2. When I do parametrize a single peptide with a single PHG residue
> and then simulate it (with zeros for the parameters in the .top
> file), the PHG turns out fine and does not get distorted. But when I
> simulate many PHG containing peptides, say 10 of them stacked on top
> of and besides each other, I observe that some of the PHGs get
> distorted while others don't. By distorted, I especially mean the
> Halpha-Calpha-aromaticC angle, which gets distorted to almost 90 from
> its normal value of about 109. Whatever the cause for this distortion
> is, why is it affecting only some PHG residues and not others?

It's a stochastic system... run it for long enough and you'll see weird
stuff everywhere.

Mark