[gmx-users] Dihedral angle PCA

[baptista at itqb.unl.pt]

On Fri, 12 Apr 2013, Thomas Evangelidis wrote:

> On 12 April 2013 07:51, anu chandra <anu80125 at gmail.com> wrote:
>
>> Hi David,
>>
>> Thanks for the reply. I have not tried yet. Since I didn’t find query about
>> the dihedral PCA in the mail list, I thought of confirm about the steps
>> mentioned in the web site.
>>
>>
>> Regarding the use of dihedral PCA, the protein with which I am working
>> behave differently from others. From literature, this protein shows minimal
>> backbone conformational changes during ligand binding and presumed that
>> there is high degree of side-chain conformational changes during binding to
>> ligand. This made me to think about dihedral PCA for get some information
>> about the ligand binding. What is your opinion in this regard?
>>
>>
> Why not Cartesian PCA restricted to the protein backbone ? dPCA is
> preferred in cases of peptides or intrinsically disordered proteins.

Even for very flexible systems, I would say that your choice between dPCA
and cPCA should depend on what you want, since your similarity metric
should match as close as possible the property you want to measure.

Since the underlying conformational spaces of dPCA and cPCA are generally 
quite distorted relative to each other, they can give quite different 
distances between conformations. For example, you can have two conformers 
that look very different in 3D space due to a series of very small 
cumulative dihedral differences, meaning that they are very distant in 
cPCA space but very close in dPCA space. Conversely, you can have two 
conformers that look very similar in 3D space due to a few very different 
dihedrals which compensate each other, meaning that they are very close in 
cPCA space but very distant in dPCA space.

So, if you want to do a detailed analysis of conformational kinetics, dPCA
is probably a good choice, because it better retains the kinetic
contiguity of torsional states. But if you are interested in molecular
recognition (ligand-receptor interactions, etc), you should probably
choose cPCA, because it better maps the overall 3D shape (assuming you can
devise a good reference structure for fit). Anyway, you have to think
about your problem and decide what better suits it.

You can see some examples and a discussion of these issues (including
using other metrics) in http://dx.doi.org/10.1021/jp902991u.

There is a very simple test you can do (check the paper): take all pairs
of simulated conformations and compute their distance in dPCA space and in
cPCA space (using all dimensions, not just the selected PCs), and then
make a scatter plot of one against the other. If the two distances are
poorly correlated, dPCA and cPCA can give quite different results.

Best,
Antonio

>
>
>> Is there any other sort of analyses that can help me to quantify such
>> side-chain conformational changes and also to locate the regions in
>> protein, which have  high degree of side-chain conformational changes
>> during ligand binding?
>>
>> Waiting for your valuable reply.
>>
>> Thanks and regards
>> Anu
>>
>>
>>
> --
> gmx-users mailing list    gmx-users at gromacs.org
> http://lists.gromacs.org/mailman/listinfo/gmx-users
> * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
> * Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org.
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
--
Antonio M. Baptista
Instituto de Tecnologia Quimica e Biologica, Universidade Nova de Lisboa
Av. da Republica - EAN, 2780-157 Oeiras, Portugal
phone: +351-214469619         email: baptista at itqb.unl.pt
fax:   +351-214411277         WWW:   http://www.itqb.unl.pt/~baptista
--------------------------------------------------------------------------