I am having some problems building ligand parameters with elbow. I have an accurate reference structure, but elbow seems to "regularize" the coordinates in a way the distorts the structure. Is there a way to disable this? With --opt, QM optimization should hopefully remove distortions, but it is not doing a very good job for this molecule. So, I decided to optimize using Gaussian, and a better basis set. It seems that the Gaussian interface is still fairly simple, and it seems that you have to edit the Gaussian.csh template. So, I did that, but got the "Gaussian does not seem to be installed" message. Unfortunately, elbow runs the Gaussian script with output to /dev/null, so there is no way to see any error messages. Maybe that should be done only when the --quiet flag is given. I edited the Python source to let errors go to the tty, and found that the problem is in the way the script is invoked using "csh -c %s.csh". That only works if "." is in your $PATH, which is often the case. It also invokes an extra instance of csh, because it is invoked as a system command. The correct way to run a shell script is to leave out the '-c'. Elbow makes the script executable, so an even better approach is to just leave out "csh", and do "os.system('./%s.csh' % project)". Then, the script can be in sh/bash/etc. Maybe the was the intent of using '-c'? Thanks, Joe Krahn
Joe On 11/4/09 9:35 AM, Joe Krahn wrote:
I have an accurate reference structure, but elbow seems to "regularize" the coordinates in a way the distorts the structure. Is there a way to disable this?
eLBOW will always run a simple optimisation unless you specify the final geometry you like using phenix.elbow --final-geometry accurate.pdb Naturally, you are smart to use a better input format thus phenix.elbow filename.sdf --final-geometry accurate.pdb
With --opt, QM optimization should hopefully remove distortions, but it is not doing a very good job for this molecule. So, I decided to optimize using Gaussian, and a better basis set.
Good thinking.
It seems that the Gaussian interface is still fairly simple, and it seems that you have to edit the Gaussian.csh template. So, I did that, but got the "Gaussian does not seem to be installed" message. Unfortunately, elbow runs the Gaussian script with output to /dev/null, so there is no way to see any error messages. Maybe that should be done only when the --quiet flag is given.
Maybe I'll implement this with a --verbose flag.
I edited the Python source to let errors go to the tty, and found that the problem is in the way the script is invoked using "csh -c %s.csh". That only works if "." is in your $PATH, which is often the case. It also invokes an extra instance of csh, because it is invoked as a system command. The correct way to run a shell script is to leave out the '-c'. Elbow makes the script executable, so an even better approach is to just leave out "csh", and do "os.system('./%s.csh' % project)". Then, the script can be in sh/bash/etc. Maybe the was the intent of using '-c'? This is old code from before PHENIX had a smart module for running commands. I'll update to code.
Nigel -- Nigel W. Moriarty Building 64R0246B, Physical Biosciences Division Lawrence Berkeley National Laboratory Berkeley, CA 94720-8235 Phone : 510-486-5709 Fax : 510-486-5909 Email : [email protected] Web : CCI.LBL.gov
participants (2)
-
Joe Krahn
-
Nigel W Moriarty