AMBER Archive (2005)

Subject: Re: AMBER: seg fault error in resp

From: FyD (fyd_at_u-picardie.fr)
Date: Wed Oct 26 2005 - 11:13:05 CDT

Quoting Kateryna Miroshnychenko <kateryna_mirosh_at_ire.kharkov.ua>:

> Yes, I just changed all the occurencies of maxq=500 to maxq=2000. I had
> to do this because earlier I've had problem with max number of
> centers(more than 500).

With Piotr, we made a 52 molecules RESP fit and it worked great.

> No, in the case of simple single molecule inputs everything is working
> fine.

OK
So for me, you have an error in your input; I mean in the input involving your
different mol. orientation and/or conformation.
Check carefully, the number of sections in your input, the "nmol" value in your
input, and the number of MEPs in the espot file: They should be the same
otherwise RESP get lost and crash... I had this problem ;-)

> This problem arose when I was trying to establish how many
> reorientations of alanine should be used to obtain charges
> with accuracy of 0.001. I compared resp charges calculated for one
> conformation using two different sets of 15 reorientations. The max
> difference in q value was circa .005. Then I wanted to compare this
> charges with charges calculated using maximum possible number of
> reorientations(60 if only heavy atoms(5 in ala) are used) . Since the
> number of all atoms in ala is 10, the total number of centers is 600.
> Here the problem occured.
> May be you can suggest any clever way to select particular
> reorientations that will minimize the charge uncertanty.

To get reproducible charge values, the idea is to pick up several mol
orientations to average the charge diff. observed for one orient. over several
orient. Charge values are reproducible _because_ you know/control the mol.
orientation in the MEP computations using the rigid body reorienation algo.
implemented in RED and _because_ you report in your work _with_ the charge
values the mol. orientation used; That's it. Taking too many orien. will not
help...

Regards, Francois

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu