Jenny,
I of course would suggest that you follow Olve's advice, and use
theseus to do a maximum likelihood, simultaneous superposition of all
your structures ( http://www.theseus3d.org ). The variable bits,
like your loop, will be naturally down-weighted in a rigorous
statistical manner. Then you can look at the average structure file
that is output (_ave.pdb at the end of the filename), and the B-
factor column has the overall RMSD for each atom in there. You can
look at the full superposition (the _sup.pdb file) in rasmol or in
pymol with the 'set all_states, on' command.
However, if you really need to do the very analysis that you asked
about, the following bash script will do exactly that with theseus
(you need both awk and theseus in your executable path). It prints
out the average RMSD for the atoms you specify in the loop, after
pairwise least-squares superpositioning on all atoms other than the
loop, for all possible pairwise combinations of your pdb files. (Note
that in this script all backslashes '\' must have a carriage return
immediately after them.) You will need to change the lower and upper
values at the top of the script (inclusive for the loop you want
excluded). You invoke the script something like "karen.sh pdb1.pdb
pdb2.pdb pdb3.pdb" or "karen.sh *.pdb" to do all the .pdbs in one
directory. If you have any problems or have other specific
superpositioning issues I'm glad to help out.
Cheers,
Douglas
karen.sh
#################################
#!/bin/bash
# everything including and between lower and upper
# is excluded from the superposition
lower=40;
upper=60;
pdbs=($*);
for (( i = 0; i < ${#pdbs[@]}; ++i ))
do
for (( j = 0; j < i; ++j ))
do
name="${pdbs[i]%.*}_${pdbs[j]%.*}";
theseus -l -r ${name} -S ${lower}-${upper} ${pdbs[i]} ${pdbs[j]}\
> ${name}.log;
rmsd=$(cut -c 7-11,61-67 ${name}_ave.pdb |\
awk '{if ($1 > lo && $1 < up) {sum += $2; n++}}; END {print sum/
n}'\
lo=${lower} up=${upper});
echo "${name} rmsd = ${rmsd}";
done
done
####################################
^`^`^`^`^`^`^`^`^`^`^`^`^`^`^`^`^`^`^`^`
Douglas L. Theobald
Department of Biochemistry
Brandeis University
Waltham, MA 02454-9110
[log in to unmask]
^\
/` /^. / /\
/ / /`/ / . /`
/ / ' '
'
On May 11, 2007, at 1:58 PM, Olve Peersen wrote:
> I would highly recommend Doug Theobald's program Theseus for this -
> the pictures at www.theseus3d.org say it all. Theseus does maximum
> likely hood superimpositions of multiple structures (i.e. NOT
> pairwise against a "master" copy), and the real beauty of it is
> that you don't have to pick which residues you want to
> superimpose. Places where the whole set of structures show
> divergence are effectively down-weighted and don't contribute much
> to the final solution vs. least squares where every atom position
> has equal weight and the "bad" parts screw up the alignment of the
> "good" parts. For this, I would do a Theseus superposition of all
> the structures and then analyze the set of superimposed structures
> by whatever method you want (e.g. rmsd of variances in important
> sections of the structures).
>
> - Olve
>
> -------------------------------------------------------
> Olve Peersen
> Associate Professor
> Dept. of Biochemistry & Molecular Biology
> 1870 Campus Delivery
> Colorado State University
> Ft. Collins, CO 80523-1870
> -------------------------------------------------------
> 970.491-0433 Office (MRB 279)
> 970.491-0271 Lab (MRB 149)
> 970.491-0494 Fax
> [log in to unmask]
> -------------------------------------------------------
>
> On May 11, 2007, at 11:15 AM, Donnie Berkholz wrote:
>
>> Eleanor Dodson wrote:
>>> It is a bit clunky - you can use siperpose molecules - fit
>>> residues to
>>> fit a selected range (1-40; 60-100 say) and write out a complete
>>> fitted
>>> pdb file. Then you could use a VERY old program
>>> compar xyzin1 original.pdb xyzin2 fitted.pdb (xyzin3 another.pdb)
>>> and it will match all pairs with the same RESIDUE ID and give the
>>> RMSD
>>> distance
>>>
>>> There is documentation for it.
>>
>> There's a nice (non-CCP4) program called ProFit that does a pretty
>> nice
>> job of superimposing with a lot of flexibility.
>>
>> Thanks,
>> Donnie
>> On May 10, 2007, at 6:45 PM, Jenny wrote:
>>> Hi, All,
>>>
>>> I have a question about rmsd calculation.
>>>
>>> I have some pdbs (100 residues ) and these pdbs differ pretty
>>> much only the loop region 40-60. Is there any easy way that I can
>>> superimpose the fixed region ( 1-40,60-100) and then calculate
>>> the rmsd for the loop?I need to calculate for each pair, so if
>>> there is any script or program available to do this quickly, that
>>> would be great.
>>>
>>> Thanks.
>>>
>>> Jenny
