[Chimera-users] energy optimization of the modle from homology modeling

[Elaine Meng]

Hello,
I agree with what Tom Goddard said:

"How to make a suitable homology model for docking and what force fields to use are not questions specific to Chimera, they are questions about how to do good quality science, and they depend a great deal on the details of your system. So you aren't likely to get the answers you want on this list."

Instead, you have to use your scientific judgement and your knowledge of this protein, how similar it is to the template protein, etc.  The procheck statistics are only one thing to consider.  One reason is that they are for the whole protein, whereas you mostly care about the binding site.  Another is that they focus on stereochemical quality, mainly torsion angles, but not other descriptors such as whether the protein is well packed.

You could see what others have done in published papers, and/or ask on the Computational Chemistry List (see <http://ccl.net/cgi-bin/ccl/send_ccl_message>) and the Modeller discussion forum (see <http://salilab.org/modeller/discussion_forum.html>) if it is recommended to minimize homology-modeled structures before docking.

In complicated structures like proteins, atoms may "feel" multiple conflicting forces, and minimization will usually relieve some strains (for example, move atoms apart if they are too close together) but increase others (for example, distort the angles of the sidechains containing those atoms). That is why the statistics from procheck got worse.  It is important to realize that minimization has very little ability to fix a complicated structure with a rugged energy landscape.  It is good for cleaning up small molecules, and sometimes for cleaning up a localized area (leaving the rest of the protein unchanged) where you have made some change such as manually rotating some bonds or mutating a residue. It will not predict any significant conformational change to find the global optimum.  If you don't have any specific reason to minimize the structure, it might be better not to minimize it.  That is only my guess, and you could probably get more expert opinions by asking your question on the other discussion lists mentioned above.

If your docking doesn't take too long, you could try docking to more than conformation of the protein (more than one model). Chimera does not do molecular dynamics, although it can show MD results from other programs. However, Modeller itself does sampling to generate multiple models.  Also, you could consider using Modeller to further refine parts of the model(s) from the first Modeller calculation.  The Modeller scores should provide some indication of which models and refinements are better.
<http://www.cgl.ucsf.edu/chimera/docs/ContributedSoftware/multalignviewer/modeller.html>

I hope this helps,
Elaine
-----
Elaine C. Meng, Ph.D.                       
UCSF Computer Graphics Lab (Chimera team) and Babbitt Lab
Department of Pharmaceutical Chemistry
University of California, San Francisco

On May 14, 2013, at 9:39 AM, Conrad Huang wrote:

> Unfortunately, that is getting well beyond my understanding of molecular mechanics.  I will have to defer this to our expert when she returns.
> 
> Conrad
> 
> On 5/13/2013 7:56 PM, aixintiankong wrote:
>> Dear,
>> Thank you for your patience to answer my questions in detail.
>>     Before /minimizating/  the model, i use the procheck to check the
>> quality of the model. the residues in most favoured regions is 95.6% and
>> the disallowed regions is 0.0%, the the residues had an averaged 3D-1D
>> score>0.2 is 92.02% and the errat is 90.55.
>>     if i use the chimera to minimizate the model after 200 steps, the
>>  most favoured regions is 85.5% and the disallowed regions is 0.4%, the
>> the residues had an averaged 3D-1D score>0.2 is 99.2% and the errat is
>> 95.968.
>>     if i use the chimera to minimizate the model after 250 using the
>> ff99sb! force filed, the most favoured regions is 92.1% and the
>> disallowed regions is 0.4%, the the residues had an averaged 3D-1D
>> score>0.2 is 96.96% and the errat is 98.000.
>>     if i use the chimera to minimizate the model after 500ns the most
>> favoured regions is 90.8% and the disallowed regions is 0.4%, the the
>> residues had an averaged 3D-1D score>0.2 is 96.58% and the errat is 96.047.
>>   which one model should i select to make docking .
>>   thank you very much!
>> 
>> At 2013-05-14 00:24:03,"Conrad Huang" <conrad at cgl.ucsf.edu> wrote:
>>> Unfortunately, our local expert on this is on vacation and will be back
>>> at the end of the week, so a more definitive answer should come then.
>>> In the mean time, I'll try to answer as best as I can.
>>> 
>>> First, Homology Modeling uses Modeller and that does a bit of
>>> optimization already.  If the model structure looks reasonable, further
>>> energy optimization might be helpful.  On the other hand, if the model
>>> has extended chains because Modeller does not have appropriate templates
>>> or just looks unreasonable because the model is not well packed or has
>>> large dangling loops, then energy minimization probably will not help.
>>> 
>>> Second, I do not know the answer, but have you tried using the default
>>> paramters?
>>> 
>>> Third, conjugate gradient is much much slower than steepest descent, so
>>> it probably pays to use steepest descent to get near the minimum before
>>> starting conjugate gradient.  You can look at the reported energy values
>>> reported to decide whether the minimization is proceeding as expected.
>>> Chimera energy minimization is not highly optimized, so it may take a
>>> while to run for a large system.  If only one portion of the model is
>>> "interesting", e.g., an active site, it might be faster to select just
>>> the region of interest (plus a buffer around it) and only optimize that
>>> part.
>>> 
>>> Hope this helps a bit.  Better answers should come later in the week.
>>> 
>>> Conrad
>>> 
>>> On 5/11/2013 6:50 PM, aixintiankong wrote:
>>>>   Dear,
>>>>    i am a newer use the chimea, i have some question to consult to you,
>>>> please help me.
>>>>   Frist, after Homology modeling using the chimera modeller, should i
>>>> optimizate the modle by energy optimization or Molecular dynamic
>>>> simulations? The next step i will use the model to make docking.
>>>>   Second, i want to use chimera to minimize the structrue of my modle.
>>>>   i select Tools>Structure Editing>Minimize Structure to perform my
>>>> work, but my modle include protein and a organic molecule NAD+, so i
>>>> don't know how to select the force fileds for the minimiztion of the
>>>> modle. please help me.
>>>>   Third, if i olny want to use the conjugate gradient to perform the
>>>> minimiztion, should i only need to set the steepest descent steps to 0?
>>>>  in general, how many conjugate gradient s! teps and size should be ?
>>>>   Thank you very much!