[Chimera-users] reasonable minimization times
goddard at sonic.net
Thu Aug 30 11:54:25 PDT 2012
If you take Elaine's suggestion and delete parts of the molecule
before minimizing you can recombine the resulting minimized fragment
with the original model by opening the original molecule and copying the
coordinates from the minimized fragment with the Chimera "mcopy"
command. For example,
mcopy #0 #1 settings x
where #0 is the fragment, #1 is the original molecule and "settings x"
means copy the coordinates.
-------- Original Message --------
Subject: Re: [Chimera-users] reasonable minimization times
From: Elaine Meng
To: David Belnap
Date: 8/30/12 8:39 AM
> Hi David,
> Oh, in that case it is the Antechamber charge calculation that has gotten stuck. It is not even getting to the minimization part. Even if you were only going to use a subset of the atoms in the minimization calculation, the entire models first get hydrogens and charges assigned.
> One of the limitations of charge-assignment in Chimera is that it does not know about the Glycam force field parameters (for glycosylation residues MAN, NAG, etc.). So instead of a simple charge lookup, it is using Antechamber to calculate the charges. To make things even worse, the glycosylation residues are all bonded, so they are treated as one big molecule for the calculation. This is simply too large a system for AM1-BCC charge calculation.
> Possible workarounds:
> (a) choose Gasteiger charge calculation method instead of AM1-BCC. It is simpler and generally faster. Also, it doesn't matter what charges are assigned to those atoms anyway since you are not including them in the minimization calculation (fragment true).
> - OR -
> (b) since these are not going to be used in the minimization anyway, you could just delete them beforehand (for example, command "delete :bma,man,nag,gal,asn,fuc" or to delete all nonprotein, "delete ~ protein"). You could add them back later after minimization by opening a new copy of the original structure, deleting only the protein from that copy, combining the models, and adding bonds back in as needed (or if you are familiar with PDB text-editing, you could save PDB of minimized structure and then text-edit to add the coordinates of the parts you had deleted, copied from a PDB file of the original structure) .
> To report specific problems it is better to use Help... Report a Bug in the menu. I mean that for future reference, since in this case it is now pretty clear what is going on. It is not a bug per se, but you are running into certain limitations as explained above.
> I hope this helps,
> Elaine C. Meng, Ph.D.
> UCSF Computer Graphics Lab (Chimera team) and Babbitt Lab
> Department of Pharmaceutical Chemistry
> University of California, San Francisco
> On Aug 29, 2012, at 10:32 PM, David Belnap wrote:
>> Thank you, Elaine. The Fc domain has 9 carbohydrate molecules that are unselected--as thus should be stationary. The program seems to hang as it displays the following at the bottom of the Chimera window:
>> I used the command option you suggested and did "clear selection" before selecting the 10 residues to be minimized. I am running this on a Mac Intel laptop (dual core machine, 2.4 GHz CPU) with 4GB RAM. I have used Chimera version 1.7 (build 37003) 2012-07-26 and 1.7 (build 37262) 2012-08-29. The process "sqm" is running at 100% when "top" is run from the command line.
>> On Aug 29, 2012, at 4:57 PM, Elaine Meng wrote:
>>> Hi David,
>>> All those parameters you mention are just for the hydrogen-addition and charge-addition steps prior to minimization and should have virtually no effect on the minimization time (assuming that is the part that takes so long).
>>> Minimization in Chimera can be very slow, in part because it does not use any distance cutoff. However, your use of "spec sel fragment true" when only 8-10 residues are selected should limit the calculations significantly, to ignore all but the selected atoms. I am surprised the calculation could take that long on such a small set of atoms. How long it should take totally depends on the computer, of course, but for 8-10 residues I haven't seen it take that long. Make sure you have only those residues selected when issuing the command.
>>> In general, "conjugate gradient" steps are significantly slower than "steepest descent" steps, so you might try setting number of conjugate gradient steps to 0 ("cgsteps 0"), at least for your initial run. There are additional parameters like step size that could have an effect, but how much of an effect or even whether increased or decreased time is not easily predictable.
>>> I hope this helps,
>>> Elaine C. Meng, Ph.D.
>>> UCSF Computer Graphics Lab (Chimera team) and Babbitt Lab
>>> Department of Pharmaceutical Chemistry
>>> University of California, San Francisco
>>> On Aug 29, 2012, at 3:32 PM, David Belnap wrote:
>>>> I am trying to link the crystal structure of an Fc domain with two Fab structures. I am using the command "minimize spec sel fragment true" on the command line. When various options come up, I've selected "if alternate locations, keep only highest occupancy", "add hydrogens", "add charges", "consider each model in isolation from all others", "also consider H-bonds", and AMBER ff12SB and ANTECHAMBER to assign charges. I've had jobs run overnight that didn't finish. What is a reasonable time for about 8-10 residues to be minimized? Can you suggest any ways to speed this up?
>>>> Thanks, David
>> Chimera-users mailing list
>> Chimera-users at cgl.ucsf.edu
> Chimera-users mailing list
> Chimera-users at cgl.ucsf.edu
More information about the Chimera-users