[Biopython-dev] RMSD calculation

Christian Schaefer schaefer at rostlab.org
Tue Nov 2 09:17:49 UTC 2010


I was using the PDB superimposer once and compared it to ProFit [1] 
which does a McLachlan fitting. Both return essentially the same rmsd, 
while the implementation in Bio.PDB seems to yield higher precision.


[1] http://www.bioinf.org.uk/software/profit/

Dipl.-Bioinf. Christian Schaefer
Technical University Munich
Department for Bioinformatics
Faculty of Computer Science/I12
Boltzmannstr. 3
D-85748 Garching b. Muenchen

On 10/30/2010 01:42 AM, George Devaniranjan wrote:
> Thanks Eric and Peter,
> Your patience in answering this question is very much appreciated.
> I think Eric maybe right, I tried the RMSD calculation for several
> structures and VMD does give a lower value for them all.
> George
> Thanks once again for all of you for your answers
> On Fri, Oct 29, 2010 at 10:39 PM, Eric Talevich<eric.talevich at gmail.com>wrote:
>> On Thu, Oct 28, 2010 at 12:49 PM, George Devaniranjan<
>> devaniranjan at gmail.com>  wrote:
>>> I was wondering why there is two functions for calculating RMSD
>>> 1)in the SVDSuperimposer()
>>> 2)in PDB.Superimposer()
>>> In the code its says RMS-is RMS being calculated instead of RMSD???
>>> I ask because VMD gives a different value for RMSD to the one from
>>> Biopython
>> Hello George,
>> Here's my understanding of it:
>> 1. RMSD and "RMS distance" both mean root mean square deviation, in terms
>> of the distances in 3D space between each corresponding pair of atoms. The
>> RMSD between all atoms in two aligned structures may be different than the
>> RMSD between backbone atoms only. Or, if the two structures don't have the
>> same peptide sequence, that raises another set of issues.
>> 2. In Biopython, PDB.Superimposer internally uses SVDSuperimposer. It's a
>> simplified wrapper.
>> 3. The SVDSuperimposer module allows you to either (i) align two structures
>> in 3D space and then calculate RMSD, or (ii) just calculate RMSD without
>> spatially (re-)aligning the structures. PDB.Superimposer just does the
>> former. If the structures weren't already aligned, these can yield very
>> different values.
>> 4. There are many ways to perform a structural alignment; SVDSuperimposer
>> implements a simple one. PyMOL, VMD, ce, DALI, and other programs implement
>> more advanced methods.
>> So don't be alarmed that VMD gives you a smaller RMSD than PDB.Superimposer
>> -- it just means VMD found a better alignment between the two structures.
>> Best,
>> Eric
> _______________________________________________
> Biopython-dev mailing list
> Biopython-dev at lists.open-bio.org
> http://lists.open-bio.org/mailman/listinfo/biopython-dev

More information about the Biopython-dev mailing list