Re: [AMBER] acpype fails to run on provided tests

From: Jason Swails <>
Date: Tue, 30 Jun 2015 00:16:57 -0400

On Mon, Jun 29, 2015 at 4:27 PM, Daniel Bork <> wrote:

> Hi,
> I have been attempting to use acypye to aid in the generation of topology
> files for custom (in this case, N-glycosylated) amino acid residues.
> However, it seems to fail to run on 2/4 of the tests provided by the
> developers (namely FFF.pdb and KKK.pdb), due to problems with sqm
> convergence. Specifically, the error acpype throws is
> Error: cannot run "[directory]/amber14/bin/sqm -O -i -o sqm.out" of
> bcc() in charge.c properly, exit
> which presumably results from the sqm error

​ suspect you should use a fragment-based approach for this. A simulation
of all 3 amino acids is much bigger than it needs to be, and the charge
derivation will likely suffer as a result.​

> QMMM: Unable to achieve self consistency to the tolerances specified
> QMMM: No convergence in SCF after 1000 steps.
> QMMM: E = -0.1184E+07 DeltaE = 0.1163E-09 DeltaP = 0.2354E-13
> QMMM: Smallest DeltaE = 0.1739E-10 DeltaP = 0.1546E-06 Step = 5

> I understand that the listed value for E is quite high; however, this
> result occurred even after extensive minimization of the provided
> phenylalanine tripeptide, so it is not obvious to me what problems the
> initial structure might have.

​The convergence criteria is likely too strict. The delta E is on the
order of 1e-10 (although changes in the density matrix are ~3 orders of
magnitude worse). If you used less stringent convergence criteria, it may
help. (Note that the *smallest* difference between steps was on step 5 --
convergence was not found on this system for over 1000 steps, which
suggests to me that the SCF will never converge to the requested tolerance
given this conformation with sqm and the requested level of theory).

​Using a fragment-based approach should help this, too, and is what I
suggest trying (tutorial B5 on the GFP at may
help you get started here). But if you just want to get it working as a
learning exercise, try reducing the convergence criteria fed to antechamber
(you may have to modify the acpype code for this, I'm not sure how acpype
works in its entirety).

Additionally, on my coworker's computer, this
> test successfully runs (after taking a much longer time than sqm took to
> announce failure to converge on my machine), so perhaps this could be
> partially a hardware issue. In case it is, I am running the latest Ubuntu
> release, 14.04.

​Likely a numerical precision issue. I think your results are *close* to
the convergence tolerance. If your coworker's computer gave *slightly*
smaller delta E or P during the SCF, it may have *just* satisfied the
limits, therefore appearing to 'work'.


Jason M. Swails
Rutgers University
Postdoctoral Researcher
AMBER mailing list
Received on Mon Jun 29 2015 - 21:30:04 PDT
Custom Search