Re: [AMBER] SHAKE problem

From: Bill Ross <>
Date: Mon, 28 Dec 2020 00:08:04 -0800

Initially it was 1.01 with shake, then later 1.33 when shake was turned
back on? If the second was a typo and 1.01 is with shake consistently,
that implies the atom has an 'H' in its type (I think), but the actual
type might be non-H, maybe through atom type aliasing(?) or the
type-type potential is wrong. Job 1 is check the types and look up what
equilibrium distance is defined.

How are you measuring? I wonder if you're measuring the same bond
consistently? 1.33 would be a C-C double bond.

I've never used QM/MM.


On 12/27/20 9:26 PM, Jonathan Church wrote:
> Hello,
> I ran into a problem using the SHAKE algorithm and I was curious if anyone had seen anything similar. I have a protein with a ligand attached embedded in a lipid bilayer. I equilibrated this system classically for about 300ns using GPU’s with CUDA from AMBER16 and during that time I used SHAKE.
> After about 300ns I then switched to using QM/MM with DFTB2+D and sander MPI. After I switched to using QM/MM with SHAKE, an N-H bond of the peptide backbone of an ILE residue about 15 angstroms from the QM region elongates to 1.33 angstrom when it was 1.01 angstrom for the entire MM equilibration period. If I turn SHAKE off the peptide N-H bond will oscillate around the 1.01 angstrom value. However as soon as I turn SHAKE back on, the bond elongates back to 1.33 angstrom.
> Has anyone else seen anything like this?
> Best,
> Jon
> _______________________________________________
> AMBER mailing list
AMBER mailing list
Received on Mon Dec 28 2020 - 00:30:02 PST
Custom Search