identify a rotation centre: domain rotation

From: WENHE ZHONG
Date: 23 November 2011 14:45


Dear members,

I would like to have your ideas if there is any way to identify a rotation centre of domain in two different states using CCP4 or other program.

The situation is: the domain of the protein will rotate between two different states (depending on substrate binding) around 8 degree, and it is (nearly) clearly that the domain is rotated around a rotation centrel. So the question is how to identify this "rotation centre" in this 3D model? The ideal is to identify a region of residues in the domain which are most closed to the rotation centre.

The tool I am using right now is the "superpose" tool in CCP4 package. The output which I think mightbe uesful is:

CENTROID OF "WORKING" MOLECULE  :              157.812 152.396 -70.778
     CENTROID OF "WORKING" MOLECULE  :(fractional)  157.812 152.396 -70.778
     CENTROID OF "REFERENCE" MOLECULE:              157.251 151.877 -70.874
     CENTROID OF "REFERENCE" MOLECULE:(fractional)  157.251 151.877 -70.874
     Distance between CENTROIDS      :                0.770
     Direction cosines of vector between CENTROIDS:   0.729   0.674   0.124

I would say the "CENTROID" it mentioned above, such as (157.251 151.877 -70.874), is possibly near to the "rotation centre". I would like to have your opinion though. Thank you.

King regards,
Wenhe

----------
From: Eleanor Dodson


I guess the rotation centre is approximately the mid point between the two centroids..

But these look surprisingly similar? I would hsave expected after an 8 degree rotation there would be some greater difference..

Eleanor

----------
From: Edward A. Berry


Would this work?
Take the rot-trans operator from superpose or lsqman and express
the rotation matrix as polar coordinates of rotation axis (and angle about it).
Get the rotation axis as direction cosines, which will be a vector along
the rotation axis of the matrix. Now take the component of the
translation vector along that rotation axis, and subtract from the
translation vector, to get the component of the translation vector
which is orthogonal to rotation axis. Divide that by two and,
as a vector from the origin, it defines a point on the
rotation axis. Add any multiple of the rotation axis vector to
this and it gives another point on the rotation axis.
Test these points by operating on them with the orig
rot-trans operator, they should just move along the rotation axis
by the screw component of the rotation

----------
From: Edward A. Berry

No! the "divide by 2" part is for a 2-fold rotation- not 8*.
Sorry, and hope the O.P. didn't waste any time trying to implement this.

----------
From: Eric Pettersen

Perhaps this video would be helpful: http://www.cgl.ucsf.edu/chimera/videodoc/AlignDomains/index.html

The "axis point" shown in the reply log at the very end is what I think you want.

BTW, the video is a little out of date, in that if you get the very latest daily build the MatchMaker tool can match the smaller domains with having to delete part of the structure.

--Eric