Hi, I encountered a strange behavior, I dont understand yet. First, I replaced Modeller generated phi/psi dihedral constraints (group 25, physical.phi_psi_dihedral) in the rsr-file. In the corresponding cuser_form.c file, I get the phi and psi angle like this (in myform_eval): float phi = feat[0]; float psi = feat[1]; printf("phi=%f, psi=%f\n", phi, psi);
Then Modeller is called.
When comparing the printed phi/psi angles, I expected the very first ones (i.e. the first ones printed into the log-file, right at the start of optimization) to be the same as the phi/psi angles extracted from the *.ini file, because this is the starting point. But this isnt at all what I see.
I hope, my description was clear enough Armin Meier
On 04/23/2010 07:50 AM, Armin Meier wrote: > When comparing the printed phi/psi angles, I expected the very first > ones (i.e. the first ones printed into the log-file, right at the start > of optimization) to be the same as the phi/psi angles extracted from the > *.ini file, because this is the starting point. > But this isnt at all what I see.
It sounds like you are running automodel, in which case the .ini file is the starting point, but before the first step of optimization, the coordinates are randomly displaced (otherwise each optimization would simply give you the same results). If you want to be sure your energy function is working, you'd be better off to generate a few test restraints on a static structure and then just call selection.energy() to test it.
Ben Webb, Modeller Caretaker
participants (2)
-
Armin Meier -
Modeller Caretaker