On Wed, Jul 02, 2025, Matthew Tremblay via AMBER wrote:
>
>I am working on QM/MM project in which my system is divided into 1) a QM
>region of two adjacent amino acids, 2) a mobile MM region that contains all
>atoms within 18 Angstroms of one of the QM atoms, and 3) a frozen MM region
>consisting of the remainder of the protein and some solvent near the
>protein surface. I'm currently equilibrating the mobile MM region for 10 ps
>using sander.MPI (from AmberTools 25) while the QM region and outer MM
>region are frozen using ibelly=1. The temperature is set to 300 K with
>Andersen-type coupling (ntt=2), no pressure coupling. The simulations are
>done without PBCs and a nonbonded cutoff larger than my system size. This
>is done for 20 replicas that differ only in the geometry of the QM region,
>and the initial conformations of the MM regions are identical across
>replicas. During this equilibration step, several of the replicas end up
>with very distorted and unphysical geometries that cause the simulation to
>fail due to SHAKE (used only for hydrogen-containing bonds) giving up. All
>of the simulations display negative temperatures throughout the simulation.
>The replicas that make it through the 10 ps oscillate around -300 K, while
>the ones that fail have their temperatures become very large negative
>numbers.
Two things to try: use restraints (ntr=1, perhaps with a weight of 5-10)
rather than ibelly=1. The way degrees of freedom and temperature are
reported might indeed be quirky with QM/MM and ibelly combined. Also, try
ntt=3 (say with gamma_ln=5 or so) rather than ntt=2. The latter is very
little used and only lightly tested, (especially in combination with
ibelly). So it's worth seeing if the problem can be isolated.
...good luck...dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sun Jul 06 2025 - 11:30:02 PDT
