[Chimera-users] Docking of PDB data

[Tom Goddard]

Hi Alexander,

  The "fit models/maps in maps" tools in Chimera just do a local 
position and orientation optimization.  So the first requirement is that 
you put the model in approximately the correct place.  The fit models 
tool is trying to maximize the average density at the selected atom 
positions.  The fit map in map tool maximizes the sum of the products of 
the two maps at positions within the shown contour level.  If the map 
you are fitting into has a region with higher density than all the 
surrounding regions then no matter where you start the model it may be 
pushed into that high density part of the map.  This happens if there 
are no intervening low density regions between the initial placement of 
the model and high density area.  If this is the situation then these 
tools may simply not be useful for you.  This is a common situation.  
Perhaps no boundaries between the 3 fragment models you are trying to 
fit is apparent in the density map.  A more sophisticated kind of 
fitting would be needed than what Chimera provides.  A capability we 
would like to add to Chimera but is not available now is to place 2 of 
your 3 fragments by hand and then fit the 3rd fragment in the map while 
avoiding steric clashes with the others.  Another capability that could 
be of use to you but is not in Chimera is fitting a map in another map 
using correlation.  Instead of maximizing the sum of products of 
density, a normalization is done so only the match between variations in 
density (high density matches high, low matches low) is optimized, and 
the fit is not pushed towards higher density regions that do not have 
good matching variations in density.  Other programs that do global 
fitting usually optimize correlation.

  Here are two ways you can use Chimera that could help.  If you can 
combine your 3 fragments into a single PDB model with a variety of 
conformations (maybe you consider 5 plausible arrangements of domains).  
Then that 3-domain assembly can be fit into the map with "fit models in 
maps" and the scores for the hypothetical conformations can be compared.

  Another possibly useful tool is volume eraser.  If the asymmetric 
units in your map are not separated by low density you could use volume 
eraser (Tools / Volume Data / Volume Eraser) to erase all but one 
asymmetric unit.  That may simplify fitting.  You could go further and 
erase parts of the density for one PKC to force the fit of a fragment to 
stay in the unerased part of the density.  It sounds like you have tried 
restricting display of the density to a subregion (Volume Viewer dialog 
subregion selection).  That has a similar effect but only allows you to 
use rectangular subregions, where the eraser tool lets you erase any 
part of the map.  You may see that the fit model gets pushed partly 
outside the bounds of the map.  That is because the atoms outside the 
map do not contribute to the optimized "average density at atom 
positions".  So if the highest density of the map is right on the edge 
of the box then the model will get pushed partly outside to get best 
overlap with that high density.

  Another technique for erasing part of the density is to place some 
models in the map, use the color zone tool to color the map within a 
certain range, then use the "split volume by color zone" plugin 
(available from 
http://www.cgl.ucsf.edu/chimera/experimental/experimental.html) to get 
just the density near that placement of the model.  Then try the fitting 
tools within that more limited density.

    Tom