Thanks John,


>I've been bit by this bug, and I wrote the underlying algorithm. On the TCL
side, I've changed from throwing an error to printing a message, so that I
>can get some output, even if it's imperfect. But that's suboptimal behavior
as well, since you actually have to look at the output of your scripts to
>discover that something was wrong.



 I also would rather have an imperfect output pdb file than nothing. It
would take less time to fix the imperfect file than to manually add all the
spin labels. Your answer suggests that there is a TCL version of
addAtoms.py; what is the TCL file name and where is it?



Jeff



  _____  

From: John Kuszewski [mailto:[email protected]] 
Sent: Tuesday, February 26, 2008 6:23 PM
To: Jeff Ellena
Cc: xplor-nih at nmr.cit.nih.gov
Subject: Re: [Xplor-nih] difficulty with addAtoms



Hi Jeff,



On Feb 26, 2008, at 5:34 PM, Jeff Ellena wrote:



Is there something other than what I did above that can be tried to get
correct geometry when using addAtoms.py with XplorNIH 2.19?



Try changing the random number seed and running again.



Is there any way to get pdb and psf file output from addAtoms.py with
XplorNIH 2.19 instead of an error message even if the geometry is not
correct?



I've been bit by this bug, and I wrote the underlying algorithm. On the TCL
side, I've changed from throwing an error to printing a message, so that I
can get some output, even if it's imperfect. But that's suboptimal behavior
as well, since you actually have to look at the output of your scripts to
discover that something was wrong.



Is there a way to make the addAtoms.py geometry correction more robust? If
yes, what is the best approach?



fixupCovalentGeom applies a bunch of short energy minimization schemes,
randomly selected from a little library, until the violations are acceptable
or the number of tries exceeds a threshold. To be honest, it's a pleasant
surprise that it's as robust as it is. 



Perhaps a change that would help would be to wrap the whole thing up in a
larger loop, restarting from the original coordinates (but a new random
seed) if the last iteration got stuck. 



There are other ways of approaching this problem (running a short dynamics
trajectory instead of minimization, or randomly perturbing the starting
coordinates before re-minimizing), but this procedure is intended to be as
gentle as possible, while obtaining the desired covalent geometry. 



--JK



-------------- next part --------------
An HTML attachment was scrubbed...
URL: 
http://dcb.cit.nih.gov/pipermail/xplor-nih/attachments/20080227/727ce309/attachment.html
 

Reply via email to