[Chimera-users] about transformation matrix

Tom Goddard goddard at cgl.ucsf.edu
Tue Jul 27 17:24:34 PDT 2010

Hi Julio,

   The details of the fitting algorithm are given in the User's Guide.


You should definitely read that if you want the best understanding.  But 
the brief answers to your questions are that the convergence criteria 
looks for optimization steps of less than 0.01 grid index to decide 
convergence.  The user interface does not allow you to change the 0.01 
grid index criteria -- it could be done by editing a line of Python.  
The default settings do mask using the contour level of the fit map as 
an outer boundary.  So that level is important.


> Hi Tom,
> Thanks so much for your detailed answer.  One important point that 
> clarified things was that Chimera's rotations are relative to (0,0,0) 
> which is different from what happens in our system, in which they are 
> relative to the center of the map. So after adding the proper 
> correction for that, we are getting reasonable results.
> However, we are still wondering about possible reasons for the small 
> residual error in the rotations/shifts.  It would seem that each time 
> I click on "Fit" it does a number of iterations (fewer than the 
> maximum number) but it doesn't seem to ever converge perfectly to 0.  
> Do you know why this happens?
> Also, does Chimera mask the maps before fitting?  If not, how are the 
> corners of the map treated during rotations?
> Thank you very much,
> Julio
> On Jul 27, 2010, at 5:10 PM, Tom Goddard wrote:
>> Hi Julio,
>>   The matrixget command gives the placement of the map in Chimera world
>> coordinates.  If you simply shift or rotate the map with the mouse this
>> changes.  That is not what you want.  You want the placement of the map
>> that was fit relative to the map it was fit into.  That is what the Fit
>> in Map dialog reports in the Reply Log.  Use that.
>>   For transformations expressed at 3 by 4 matrices Chimera does the
>> rotation first (about 0,0,0) then the translation.
>>     Tom
>>> Hi,
>>> I am doing fitting of two maps and am getting different values of 
>>> the transformation parameters: the ones given in the window where I 
>>> specify the options (these seem to be correct), and those given by 
>>> using the command "matrixget" --- these give a very large 
>>> translation vector, and moreover, I don't see any relation between 
>>> that vector and the correct translation.
>>> Also, could you tell me what is the order used in Chimera: rotation, 
>>> then translation?  or  translation, then rotation?
>>> Thank you very much.
>>> Julio Kovacs

More information about the Chimera-users mailing list