As someone said - this is quite hard unless you have a very long helix -
any ragged end bits can dominate the fit of one feature to another.
In your case I think I would use SSM to superpose the two similar
structures , then LSQKAB to fit any feature to its related one using the
original molecule, plus the second one after the SSM overlap.
LSQKAB will give you the relative rotation of any feature to its partner
- look the the polar angles to get a estimate of rotation, and the
translation to find how far apart the 2 features are.
This is different to getting the direction of the helix. Centre of mas
is easy LSQKAB gives you that, but the vector is easisest found with a
bit of arithmetic.
Find COM of residues 1-3 say and COM of residues n to n-3,
vector connects these two COMs - direction cosines are
xv/(sqrt(xv*xv +yv*yv +zv*zv) yv/(sqrt(xv*xv +yv*yv +zv*zv ..
length is a function of number of residues
The $CLIBD/fraglib/theor-helix-70.pdb suggests ~ 14.8A per 10 residues..
Eleanor
Phil Evans wrote:
> The problem with the inertial matrix approach is that it is very sensitive to end effects on the helix, ie a helix is not a perfect cylinder. So superimposing an "ideal" helix is more reliable
> Phil
>
>
> On 17 Aug 2010, at 10:17, Francois Berenger wrote:
>
>> Hello,
>>
>> Is there some C or C++ code out there doing what you described in 1).
>>
>> If not, is there a very detailed explanation of this procedure somewhere, detailed enough in order to implement it (just getting
>> the best fit vector and its "length", no other parameters)?
>>
>> Thanks a lot,
>> Francois.
>>
>> Tom Oldfield wrote:
>>> Yuan SHANG
>>> 1) DIY
>>> The way that has been used is to calculate the inertia tensor matrix for helix (or
>>> any other secondary structure element). You can chose backbone atoms or just
>>> the CA atoms. Then calculate the eigen vectors and values from this and the largest
>>> eigen vector will be the best fit vector to the helix - and its lambda will define its
>>> "length". For a strand or sheet you can use this method too.
>>> This was the standard way from molecular simulation work to look at
>>> simplified dynamics of proteins.
>>> 2) The program Squid
>>> http://www.ebi.ac.uk/~oldfield/squid/ (1992, 1998)
>>> has lots of different analysis methods for proteins including calculating
>>> vectors for helices, the angles between helices (torsion/distance/opening)
>>> and other things.
>>> You only problem is that it is very old (1988) and written in Fortran and requires
>>> a little effort to install - sorry - I no longer support it. There is a pre
>>> compiled linux-32 bit
>>> version and I still do all my structure analysis with it.
>>> http://www.ebi.ac.uk/~oldfield/xsquid - though this requires installation
>>> data too.
>>> Tom
>>>> Fitting a helix is not trivial.
>>>>
>>>> If you have access to windows and mathematica, then you might try helfit. (Otherwise, you could implement the algorithm yourself and then share your code with the rest of us ;-)
>>>>
>>>>
>>>> http://dx.doi.org/10.1016/j.compbiolchem.2008.03.012
>>>>
>>>>
>>>> James
>>>>
>>>>
>>>> On Aug 15, 2010, at 12:29 AM, 商元 wrote:
>>>>
>>>>> Dear all,
>>>>> I want to compare the conformational change of two similar structures, using one alpha helix as the reference. Then, how can I get a vector that can represent both the position and direction of the helix? Is there any well-known software can do this?
>>>>> Or, should I build a cylinder model, with parameters [radius,bottom center(x1,y1,z1),top center(x1,y2,z2)], using the coordinates of C,C(alpha) and N to fit these parameters?
>>>>> Thanks for any suggestions
>>>>>
>>>>> Regards,
>>>>> Yuan SHANG
|