[Chimera-users] Minimize with fixed atoms

[Tom Goddard]

Hi Darrell,

   This is certainly a bug since the update interval is only supposed to 
effect how often the display is updated during the minimization.  I 
tried your settings on PDB 1a0m with chain A selected and observed the 
same problem that with update interval 1 steps are all 0 with no change 
in energy or gradient, while update interval 10 gives a significant 
change in energy and a small step.  The Chimera daily build has the same 
bug.  I file a report and put your name on it so you get notified when 
it is fixed.

     Tom


> Hi Tom (and Eric),
>
> Thanks for the quick responses!
>
> I'm actually trying to do this demonstration for a class I'll be teaching soon. Because I'll only be demonstrating instead of doing production work, I turned the settings for the minimization down:
>
> Steepest Descent Steps: 10
> Steepest Descent Step size (A): 0.02
> Conjugate Gradient Steps: 0
> Conjugate Gradient Step size (A): 0.02
> Update interval: 1
> Fixed atoms: unselected
>
> What I've figured out is that I'm getting trouble by using the "Update interval" at "1". In my cleverness, I wanted to recalculate the clash (using the "after relative motions (until dialog closed)") after every step of minimization to show the clashes going away after just a few steps of minimization.
>
> Everything works as expected when I put the "Update interval" at "5" and "10". Intervals of "1", "2", "3", and "4" don't work (RMSD = 0.000000 at each report). I haven't tried other interval settings.
>
> So it looks like I'll just use a higher interval setting for my class.
>
> BTW, thanks for the updates to the minimization engine. I really appreciate the prep options that check for alternate positions, uncomplexed ions, etc. And for the options to easily set the hydrogenation state of various residues and the charges of hetero atoms.
>
> Chimera is great!
>
> (BTW, this is Chimera 1.6.1 (build 35849) for 64-bit Mac OS X.)
>
> Cheers,
> Darrell
>
>
> On Apr 2, 2012, at 9:49 PM, Tom Goddard wrote:
>
> Hi Darrell,
>
>    I've often seen Chimera's MMTK energy minimization go nowhere as you
> observe.  I think basically it is taking miniscule steps, essentially
> zero, because there is a bad clash or extremely long bond.  In fact
> whatever is causing the energy to be astronomical is what is probably
> what is causing the minimizer not to take any step bigger than 0.  I
> don't know a good way to see where the high energy is coming from.
> Maybe you could try minimizing some very small set of the atoms with the
> rest fixed.  By trying minimizing different movable sets of atoms you
> could see where the high energy is coming from.  I'm not sure this will
> work, because it could be that the step size is forced to be tiny even
> if the high energy is caused by fixed atoms.  You'd have to experiment
> to figure that out.  I'll be interested to hear what you figure out.
>
>    Another idea that could help you figure this out -- the Chimera
> "minimize" command has a new "fragment true|false" option that allows
> you to minimize just part of a molecule.  It is different from holding
> atoms fixed.  With fixed atoms as you have tried, the fixed atoms still
> contribute to the force field.  With "fragment true" and the minimize
> command only the specified atoms contribute to the force field, the
> others being completely ignored.  This fragment option is not available
> in the minimize dialog.  It was added in Chimera 1.6.
>
>      http://www.cgl.ucsf.edu/chimera/docs/UsersGuide/midas/minimize.html
>
>    I see it would be nice to have Chimera be able to color your atoms
> and bonds in a way that reflects the energy terms for involving each
> atom and bond.  Not sure if that info is available in the dynamics
> library Chimera uses, MMTK.  That would allow you to quickly spot the
> problem.
>
>      Tom
>
>
> Hi there,
>
> I'm trying to minimize a protein after mutating a residue with the Rotamers panel. I first do a Check Clashes and select the clashing atoms. Then I expand the selection using Zone to<5 A. I use these selected atoms as the non-fixed atoms and the unselected atoms as the fixed ones. When I then run the minimization with the appropriate settings, I get an RMSD of 0.000 A for each step, even though the step size is 0.02 A. Nothing moves. Same thing if I have no fixed atoms.
>
> Am I in a local minima? Is that why nothing moves?
>
> The energy is very high (from the Reply Log):
>
> Potential energy: 222274199.525130, Gradient norm: 833984762.552943
> Updated 6029 atoms.  RMSD: 0.000000
> Finished 1 of 5 steepest descent minimization steps
>
> Any ideas? I do have some hetero atoms...
>
> Thanks,
> Darrell
>
>