[AMBER] Input case causing NaN in sander

From: David Case <david.case.rutgers.edu>
Date: Wed, 31 May 2017 22:10:34 -0400

On Mon, May 29, 2017, Ake Sandgren wrote:

> I can't see any obvious problems using checkoverlap, but on the other
> hand i'm not quite sure if i'm using it correctly.
> The input files are available from
> https://web-upload.its.umu.se/dl/f/1496089012_27796/inf-nan-problem.tar.bz2

Thanks. Here's how I went about debugging the problem:

1. Always try a short run (say a one-step minimization) setting ntpr=1.
That way you get the most information. With your files, that gives this
as the mdinfo file:

      1 NaN NaN 1.7616E+03 O 11561

 BOND = 289.3063 ANGLE = 1322.4636 DIHED = 4107.7576
 VDWAALS = NaN EEL = NaN HBOND = 0.0000
 1-4 VDW = 1490.2773 1-4 EEL = 21931.0867 RESTRAINT = 0.0000

The fact that the vdw and electrostatic interactions are NaN, but other things
seem OK, is a pretty good sign that two atoms (or more) are on top of each

Then, I used cpptraj

cpptraj HumanTTR_TCP_Complex_Solvated.prmtop
trajin HumanTTR_TCP_Complex_Solvated.inpcrd
checkoverlap reportfile bad-contacts.txt

Here is the bad-contacts.txt file:

1 Warning: Atoms 7730:WAT_688.O and 8558:WAT_964.O are close (0.00)
1 Warning: Atoms 7731:WAT_688.H1 and 8559:WAT_964.H1 are close (0.00)
1 Warning: Atoms 7732:WAT_688.H2 and 8560:WAT_964.H2 are close (0.00)

Clearly, waters 688 and 964 are in identical postions. I hope this
information will help your user track down the origin of the problem.


AMBER mailing list
Received on Wed May 31 2017 - 19:30:03 PDT
Custom Search