VMD-L Mailing List

From: Axel Kohlmeyer (akohlmey_at_gmail.com)
Date: Sun Jan 17 2010 - 11:54:03 CST

On Sun, 2010-01-17 at 12:11 +0100, Balazs Jojart wrote:
> Dear Users,
> I have a problem during trajectory evaluation.
> A would like to measure distance between two atoms in the trajectory.
> I know, that without deleting the selection, variables I can shock my
> computer.

well, currently you are only shocking people
reading your script because of the number of
"don't" that you have committed. there are
quite a few good template examples that explain
how to do this kind of thing, and you seem to
have started from a pretty bad one.

> Here is a part of my tcl script:
>
> mol new {../1.00M_GPCR_ions.pdb} type {pdb} first 0 last -1 step 1 waitfor 1
> mol addfile {../1.00_PROD3_CASE1.dcd} type {dcd} first -1 last -1 step
> 1 waitfor -1 0

what is the 0 at the end of this line for?

> set outfile1 [open 1.00_PROD1_Asp52OD1_NA_min.txt w]
> set nf [molinfo top get numframes]

> for {set i 1} {$i < $nf} {incr i} {
> puts "frame $i of $nf"
> set selatoms1 [[atomselect top "protein and resid 52 and name OD1 or
> resname NA and resid 1"] get index]

ouch!!! this will produce a memory leak in _any_ case.
the reason being, that the atomselect call returns the
name of the new atomselect### command being created,
but you don't store it and rather immediately use it.
thus $selatoms contains only the indices of the atoms
that you are interested in.

what you have to do instead, is assign the return value
of the atomselect call to a variable, and then generate
the list of indices from that and assign it to a different
variable.

> set bond1 [list [lindex $selatoms1 0] [lindex $selatoms1 1]]

this is a pointless operation. $selatoms1 has
the exact same content as $bond1 will have.

> set d1 [measure bond $bond1 frame $i]

> $selatoms1 delete

this is where you get the error. since $selatoms contains a
list of atom indices and not the name of an atomselect###
command, it will try to execute those numbers, but there is
no command of that name.

> $bond1 delete

this will give you the same error. have a proper look at a
tcl documentation or tutorial or both. variables are not
deleted this way, but rather using the unset command. it
is not really needed here anyways.

> set selatoms2 [[atomselect top "protein and resid 52 and name OD1 or
> resname NA and resid 2"] get index]
> set bond2 [list [lindex $selatoms2 0] [lindex $selatoms2 1]]
> set d2 [measure bond $bond2 frame $i]
> ........
> }
>
> Unfortunatelly, I obtain the following error:

this is not unfortunate but deserved. there are a bunch of
issues. first of all, there is no point in creating the
atom selections in every step. it is much better to create
them once outside the loop and then just move them ahead
or update them. since you want to refer to fixed atoms,
there is not even a need to redo the selection. the order
of their indexes will never change. so place the following
_outside_ the loop.

set sel1 [atomselect top "protein and resid 52 and name OD1 or resname
NA and resid 1"]
set bond1 [$sel1 list]

set sel2 [atomselect top "protein and resid 52 and name OD1 or resname
NA and resid 2"]
set bond2 [$sel2 list]

$sel1 delete
$sel2 delete

and then you can use $bond1 and $bond2 with the measure command, just
as you did.

also, you should rethink the syntax of your selection text. it may not
do exactly what you expect. i would place parenthesis to define proper
precedence. even if it currently works, it may fail in unexpected ways
later on.

> > frame 1 of 7151
> > invalid command name "790 4926"
> > Info) VMD for LINUXAMD64, version 1.8.7 (August 1, 2009)
> > Info) Exiting normally.
> Where is the problem in my script, and how can I solve it?

there are multiple problems, and the best solution is to study
existing good scripts. there are examples in the VMD user's guide,
tutorials and in the script library. i also suggest checking out
the tcl tutorial.

cheers,
   axel.
> Any help will be appreciated!
> Thank you for your help in advance,
> Balazs
> 

-- 
Dr. Axel Kohlmeyer  akohlmey_at_gmail.com 
Institute for Computational Molecular Science
College of Science and Technology
Temple University, Philadelphia PA, USA.