Hi Ed,
Pierre-Yves and I are now testing the fix...
We set-up three different tests including all datasets (for all ~150
residues). After the first round of prolate completed, we will know if
residue 44 is still associated with a ts of ~20 ns in model 'm8' or if
this model is eliminated...
1.
Start from scratch with full_analysis.py. The local_tm run will be done
and, after, the prolate run.
2.
Use the old local_tm directory (from analysis before the bug report of
yesterday) and only make the prolate run...
3.
Use the old prolate directory (from analysis before the bug report of
yesterday) including the init and round_1 sub-directories and only make
round 2.
We will have to wait to see if everything looks normal.
However, there seems to be a problem, already. For test #2, we get the
following error in the inital round of optimisation for the diffusion
tensor (init round):
==============
relax> diffusion_tensor.init(params=(1e-08, 0, 0, 0), time_scale=1.0,
d_scale=1.0, angle_units='deg', param_types=0, spheroid_type='prolate',
fixed=False)
relax> fix(element='all_spins', fixed=True)
relax> grid_search(lower=None, upper=None, inc=11, constraints=True,
verbosity=1)
The diffusion tensor parameters together with the model-free parameters
for all spins will be used.
Unconstrained grid search size:
3483241635926760276239670296061783033856385373964443346380512727207030807448244020315627244532223540371618455651157976598330893796730564039193972812374440994792715667315064916696072753116546554132476003724628552320667084135787252303417928004684992625923335554734448200360162656232011377981776056433588143239686879331L
(constraints may decrease this size).
RelaxError: A grid search of size
3483241635926760276239670296061783033856385373964443346380512727207030807448244020315627244532223540371618455651157976598330893796730564039193972812374440994792715667315064916696072753116546554132476003724628552320667084135787252303417928004684992625923335554734448200360162656232011377981776056433588143239686879331L
is too large.
==============
This problem also arises when not making the change at line 303
(self.model_selection(modsel_pipe='final', dir=self.base_dir + 'aic')).
So, my guess is that this problem might be due to changes in r8999.
Regards,
Séb :)
Edward d'Auvergne wrote:
Actually, this will cause a problem. I've closely looked at the
script, and it appears that your suggestion is correct:
I just tried something using the files in the bug report, as well as with
the up-to-date 1.3 repository (r9001)...
I modified line 303 of the script from:
self.model_selection(modsel_pipe='final',
dir=self.base_dir + 'aic')
to:
self.model_selection(modsel_pipe='aic',
dir=self.base_dir + 'aic')
With this change, model 'm8' is eliminated and model 'm1' is selected
instead.
This data pipe is for AIC model selection and the results go into the
'aic' directory. This is followed by the 'opt' data pipe for
diffusion tensor optimisation. So I'll make the change to the
repository. But could you check if all works with the changes that
were made to the 1.3 line? I think these changes were logical, but
there is a regression causing a system test to fail.
Cheers,
Edward
--
Sébastien Morin
PhD Student
S. Gagné NMR Laboratory
Université Laval & PROTEO
Québec, Canada
Re: [bug #13259] full analysis-high Te values