Hi Charles,
thanks again for you valuable input. Especially your note on adding the
config actions is really important for the goal of my calculations, thank
you for pointing that out!
I still have bad news however: I tried using the current version of
Xplor-NIH (3.2) on two different machines (one running Fedora 31 on
Linux-Kernel 5.8.18, the other openSUSE 15.1, Linux-Kernel 4.12.14). In both
cases I get a termination with the same error as before updating. I would
really like to troubleshoot this, but as of now I cannot figure out what is
happening or how to reproduce failure/success in these cases. I will update
you as soon as I find the time to dig deeper into the IVM setup, and of
course, all hints are highly appreciated.
All the best,
Till
On Tue, 05 Jan 2021 16:23:02 -0500
Charles Schwieters <[email protected]> wrote:
Hi Till--
> 2) Do the calcXTensor procedure separately, as you are doing now,
but
> then restrain the positions to be the same during the remainder of
the
> calculation. This may be better, since there is no reason the
> alignment tensors will be the same.
As of now, all PCS sets are based on the same (in-silico) Terbium
atom.
So that should already do the trick?
Indeed- this is what I intended.
> This indicates a problem with the IVM setup. I *might* be able to
figure
> out the problem if you send the complete script.
I created and attached an archive, with which I can reproduce the
behaviour. It also includes some needed modifications. Hopefully
there
is no problem running this, otherwise I could try sending it
differently.
I received your .zip file successfully, but I suggest that you send
such information directly to my email, as this is a public mailing
list.
I ran the script run.py successfully. One thing which bothers me, is
that you seem to be using an intermediate release of Xplor-NIH. I
suggest you try the current release, version 3.2.
A few things:
1) Currently, the alignment tensor Da and Rh are fixed at the values
from the initial calls to calcXTensor. You might add the bits from
eginput/fold.py in the Xplor-NIH distribution I included below to
optimize these parameters during final minimization. The calls to
setFreedom should be done before the calls to
protocol.torsionTopology
or protocol.cartesianTopolgy. The file eginput/refine.py takes a
more
aggressive approach to tensor parameter optimization allowing Da and
Rh to float throughout, using calcTensor at each simulated annealing
temperature.
2) You might wish to increase the PCS fore constant.
3) The script runs somewhat slowly due to the PyPot term you
wrote. Addressing bottlenecks in that term will make the calculation
faster, but this you probably understand well.
The crash you reported looks like a problem with the IVM setup of
pseudoatom topology.
dyn.addConfigAction( IVMAction("""
for m in media.values():
m.setFreedom("fixDa, fixRh") #fix tensor Rh, Da, vary
orientation
""") )
minc.addConfigAction( IVMAction("""
for m in media.values():
m.setFreedom("varyDa, varyRh") #vary tensor Rh, Da, vary
orientation
""") )
########################################################################
To unsubscribe from the XPLOR-NIH list, click the following link:
http://list.nih.gov/cgi-bin/wa.exe?SUBED1=XPLOR-NIH&A=1
