I'm trying to do a loop refinement calculation with the dopehr_loopmodelclass, modeller 9v4. Here's the python script I'm using: from modeller import * from modeller.automodel import * log.verbose() # request verbose output env = environ() # directories for input atom files env.io.atom_files_directory = ['./', 'PDB/'] a = dopehr_loopmodel(env, alnfile='TTN.ali', knowns='2g8gOligo', sequence='1s58', assess_methods=(assess.DOPE, assess.GA341)) a.starting_model = 1 a.ending_model = 1 a.md_level = None # No refinement of model a.loop.starting_model = 1 a.loop.ending_model = 1 a.loop.md_level = refine.fast a.make()
After a few hours of running, it stop. Here's the last part of the log: ... randomi_498_> Atoms,selected atoms,random_seed,amplitude: 36972 36056 1 5.0000 randomi_496_> Amplitude is > 0; randomization is done. preppdf_458W> Both Lennard-Jones and statistical potential terms selected.
Dynamically allocated memory at amaxrestraints [B,KiB,MiB]: 327553569 319876.531 312.379
Dynamically allocated memory at amaxrestraints [B,KiB,MiB]: 554709537 541708.562 529.012 preppdf_458W> Both Lennard-Jones and statistical potential terms selected. check_inf__E> Atom 31754 has out-of-range coordinates (usually infinity). The objective function can thus not be calculated.
>> Summary of successfully produced models: Filename molpdf DOPE score GA341 score ---------------------------------------------------------------------- 1s58.B99990001.pdb 75059.09375 -361379.46875 0.78559
>> Summary of failed loop models: 1s58.BL00010001.pdb check_inf__E> Atom 31754 has out-of-range coordinates (usually infinity). The objective function can thus not be calculated.
Dynamically allocated memory at finish [B,KiB,MiB]: 549216697 536344.438 523.774 Starting time : 2008/10/24 10:37:24 Closing time : 2008/10/24 12:51:48 Total CPU time [seconds] : 8018.72
The model part went OK, but when it tried to do the loop refinement it went crazy (the coordinates in1s58.BL00010001.pdb are all messed up).
I'm also testing the other two classes (loopmodel and dope_loopmodel) and they've been running for a long while now, so I don't know if they'll work or not. I'd just like to know if I'm doing something wrong or I just ran into some kind of glitch or bug in the dopehr algorithm.
Thanks!
Mauricio Carrillo Tripp wrote: > I'm trying to do a loop refinement calculation with the dopehr_loopmodel > class, modeller 9v4. > Here's the python script I'm using: ...
You're only building one loop model. I'd recommend building several, so that you can sample the conformational space more effectively. This also allows for the possibility that some of the models are not produced due to optimizer failure, as you have discovered.
... > check_inf__E> Atom 31754 has out-of-range coordinates (usually infinity). > The objective function can thus not be calculated.
Most likely the loop conformation is distorted in such a way that the optimizer cannot recover. In the case of GB/SA (used by dope_loopmodel) this may result in very large forces, thereby causing very distorted models. The older loopmodel class is more stable because it does not use GB/SA.
Otherwise, it looks like what you are doing is fine.
Ben Webb, Modeller Caretaker
Thank you, my loop refinement calculation has ended successfully now, however...
You're only building one loop model. I'd recommend building several, so > that you can sample the conformational space more effectively. This also > allows for the possibility that some of the models are not produced due > to optimizer failure, as you have discovered. > > OK, so after I get a bunch of models with a refined loop, how can I tell which one is better than the others?
Mauricio Carrillo Tripp wrote: > OK, so after I get a bunch of models with a refined loop, how can I tell > which one is better > than the others?
Well, you can't be sure, of course, but one way is to assess them all with DOPE (or some other assessment score) and take the one (or cluster) with lowest DOPE score.
Ben Webb, Modeller Caretaker
participants (2)
-
Mauricio Carrillo Tripp -
Modeller Caretaker