Author: tlinnet
Date: Tue Aug 12 15:23:57 2014
New Revision: 25000
URL: http://svn.gna.org/viewcvs/relax?rev=25000&view=rev
Log:
Added keyword to relax_disp auto analysis, if R2eff values should be
optimised.
Here optimisation means minimisation and Monte Carlo simulations of the error.
sr #3135(https://gna.org/support/?3135): Optimisation of the R1 relaxation
rate for the off-resonance R1rho relaxation dispersion models.
Modified:
branches/R1_fitting/auto_analyses/relax_disp.py
Modified: branches/R1_fitting/auto_analyses/relax_disp.py
URL:
http://svn.gna.org/viewcvs/relax/branches/R1_fitting/auto_analyses/relax_disp.py?rev=25000&r1=24999&r2=25000&view=diff
==============================================================================
--- branches/R1_fitting/auto_analyses/relax_disp.py (original)
+++ branches/R1_fitting/auto_analyses/relax_disp.py Tue Aug 12 15:23:57
2014
@@ -51,38 +51,39 @@
opt_func_tol = 1e-25
opt_max_iterations = int(1e7)
- def __init__(self, pipe_name=None, pipe_bundle=None, results_dir=None,
models=[MODEL_R2EFF], grid_inc=11, mc_sim_num=500, exp_mc_sim_num=None,
modsel='AIC', pre_run_dir=None, insignificance=0.0, numeric_only=False,
mc_sim_all_models=False, eliminate=True, set_grid_r20=False):
+ def __init__(self, pipe_name=None, pipe_bundle=None, results_dir=None,
models=[MODEL_R2EFF], grid_inc=11, mc_sim_num=500, exp_mc_sim_num=None,
modsel='AIC', pre_run_dir=None, optimise_pre_run_r2eff=True,
insignificance=0.0, numeric_only=False, mc_sim_all_models=False,
eliminate=True, set_grid_r20=False):
"""Perform a full relaxation dispersion analysis for the given list
of models.
- @keyword pipe_name: The name of the data pipe containing all
of the data for the analysis.
- @type pipe_name: str
- @keyword pipe_bundle: The data pipe bundle to associate all
spawned data pipes with.
- @type pipe_bundle: str
- @keyword results_dir: The directory where results files are
saved.
- @type results_dir: str
- @keyword models: The list of relaxation dispersion models
to optimise.
- @type models: list of str
- @keyword grid_inc: Number of grid search increments. If
set to None, then the grid search will be turned off and the default
parameter values will be used instead.
- @type grid_inc: int or None
- @keyword mc_sim_num: The number of Monte Carlo simulations to
be used for error analysis at the end of the analysis.
- @type mc_sim_num: int
- @keyword exp_mc_sim_num: The number of Monte Carlo simulations
for the error analysis in the 'R2eff' model when exponential curves are
fitted. This defaults to the value of the mc_sim_num argument when not
given. For the 2-point fixed-time calculation for the 'R2eff' model, this
argument is ignored.
- @type exp_mc_sim_num: int or None
- @type
- @keyword modsel: The model selection technique to use in
the analysis to determine which model is the best for each spin cluster.
This can currently be one of 'AIC', 'AICc', and 'BIC'.
- @type modsel: str
- @keyword pre_run_dir: The optional directory containing the
dispersion auto-analysis results from a previous run. The optimised
parameters from these previous results will be used as the starting point for
optimisation rather than performing a grid search. This is essential for
when large spin clusters are specified, as a grid search becomes
prohibitively expensive with clusters of three or more spins. At some point
a RelaxError will occur because the grid search is impossibly large. For the
cluster specific parameters, i.e. the populations of the states and the
exchange parameters, an average value will be used as the starting point.
For all other parameters, the R20 values for each spin and magnetic field, as
well as the parameters related to the chemical shift difference dw, the
optimised values of the previous run will be directly copied.
- @type pre_run_dir: None or str
- @keyword insignificance: The R2eff/R1rho value in rad/s by which
to judge insignificance. If the maximum difference between two points on all
dispersion curves for a spin is less than this value, that spin will be
deselected. This does not affect the 'No Rex' model. Set this value to 0.0
to use all data. The value will be passed on to the
relax_disp.insignificance user function.
- @type insignificance: float
- @keyword numeric_only: The class of models to use in the model
selection. The default of False allows all dispersion models to be used in
the analysis (no exchange, the analytic models and the numeric models). The
value of True will activate a pure numeric solution - the analytic models
will be optimised, as they are very useful for replacing the grid search for
the numeric models, but the final model selection will not include them.
- @type numeric_only: bool
- @keyword mc_sim_all_models: A flag which if True will cause Monte
Carlo simulations to be performed for each individual model. Otherwise Monte
Carlo simulations will be reserved for the final model.
- @type mc_sim_all_models: bool
- @keyword eliminate: A flag which if True will enable the
elimination of failed models and failed Monte Carlo simulations through the
eliminate user function.
- @type eliminate: bool
- @keyword set_grid_r20: A flag which if True will set the grid
R20 values from the minimum R2eff values through the r20_from_min_r2eff user
function. This will speed up the grid search with a factor
GRID_INC^(Nr_spec_freq). For a CPMG experiment with two fields and standard
GRID_INC=21, the speed-up is a factor 441.
- @type set_grid_r20: bool
+ @keyword pipe_name: The name of the data pipe
containing all of the data for the analysis.
+ @type pipe_name: str
+ @keyword pipe_bundle: The data pipe bundle to
associate all spawned data pipes with.
+ @type pipe_bundle: str
+ @keyword results_dir: The directory where results
files are saved.
+ @type results_dir: str
+ @keyword models: The list of relaxation
dispersion models to optimise.
+ @type models: list of str
+ @keyword grid_inc: Number of grid search
increments. If set to None, then the grid search will be turned off and the
default parameter values will be used instead.
+ @type grid_inc: int or None
+ @keyword mc_sim_num: The number of Monte Carlo
simulations to be used for error analysis at the end of the analysis.
+ @type mc_sim_num: int
+ @keyword exp_mc_sim_num: The number of Monte Carlo
simulations for the error analysis in the 'R2eff' model when exponential
curves are fitted. This defaults to the value of the mc_sim_num argument
when not given. For the 2-point fixed-time calculation for the 'R2eff'
model, this argument is ignored.
+ @type exp_mc_sim_num: int or None
+ @keyword modsel: The model selection technique to
use in the analysis to determine which model is the best for each spin
cluster. This can currently be one of 'AIC', 'AICc', and 'BIC'.
+ @type modsel: str
+ @keyword pre_run_dir: The optional directory
containing the dispersion auto-analysis results from a previous run. The
optimised parameters from these previous results will be used as the starting
point for optimisation rather than performing a grid search. This is
essential for when large spin clusters are specified, as a grid search
becomes prohibitively expensive with clusters of three or more spins. At
some point a RelaxError will occur because the grid search is impossibly
large. For the cluster specific parameters, i.e. the populations of the
states and the exchange parameters, an average value will be used as the
starting point. For all other parameters, the R20 values for each spin and
magnetic field, as well as the parameters related to the chemical shift
difference dw, the optimised values of the previous run will be directly
copied.
+ @type pre_run_dir: None or str
+ @keyword optimise_pre_run_r2eff: Flag to specify if the read
previous R2eff results should be optimised. For R1rho models where the error
of R2eff values are determined by Monte-Carlo simulations, it can be valuable
to make an initial R2eff run with a high number of Monte-Carlo simulations.
Any subsequent model analysis can then be based on these R2eff values,
without optimising the R2eff values.
+ @type optimise_pre_run_r2eff: bool
+ @keyword insignificance: The R2eff/R1rho value in rad/s
by which to judge insignificance. If the maximum difference between two
points on all dispersion curves for a spin is less than this value, that spin
will be deselected. This does not affect the 'No Rex' model. Set this value
to 0.0 to use all data. The value will be passed on to the
relax_disp.insignificance user function.
+ @type insignificance: float
+ @keyword numeric_only: The class of models to use in
the model selection. The default of False allows all dispersion models to be
used in the analysis (no exchange, the analytic models and the numeric
models). The value of True will activate a pure numeric solution - the
analytic models will be optimised, as they are very useful for replacing the
grid search for the numeric models, but the final model selection will not
include them.
+ @type numeric_only: bool
+ @keyword mc_sim_all_models: A flag which if True will cause
Monte Carlo simulations to be performed for each individual model. Otherwise
Monte Carlo simulations will be reserved for the final model.
+ @type mc_sim_all_models: bool
+ @keyword eliminate: A flag which if True will enable
the elimination of failed models and failed Monte Carlo simulations through
the eliminate user function.
+ @type eliminate: bool
+ @keyword set_grid_r20: A flag which if True will set
the grid R20 values from the minimum R2eff values through the
r20_from_min_r2eff user function. This will speed up the grid search with a
factor GRID_INC^(Nr_spec_freq). For a CPMG experiment with two fields and
standard GRID_INC=21, the speed-up is a factor 441.
+ @type set_grid_r20: bool
"""
# Printout.
@@ -104,6 +105,7 @@
self.exp_mc_sim_num = exp_mc_sim_num
self.modsel = modsel
self.pre_run_dir = pre_run_dir
+ self.optimise_pre_run_r2eff = optimise_pre_run_r2eff
self.insignificance = insignificance
self.set_grid_r20 = set_grid_r20
self.numeric_only = numeric_only
@@ -481,14 +483,21 @@
self.interpreter.value.set(param=param, index=None)
# Minimise.
- self.interpreter.minimise.execute('simplex',
func_tol=self.opt_func_tol, max_iter=self.opt_max_iterations,
constraints=True)
+ if model == MODEL_R2EFF:
+ if self.optimise_pre_run_r2eff:
+ self.interpreter.minimise.execute('simplex',
func_tol=self.opt_func_tol, max_iter=self.opt_max_iterations,
constraints=True)
+ else:
+ pass
+ else:
+ self.interpreter.minimise.execute('simplex',
func_tol=self.opt_func_tol, max_iter=self.opt_max_iterations,
constraints=True)
+
# Model elimination.
if self.eliminate:
self.interpreter.eliminate()
# Monte Carlo simulations.
- if self.mc_sim_all_models or len(self.models) < 2 or model ==
MODEL_R2EFF:
+ if self.mc_sim_all_models or len(self.models) < 2 or (model ==
MODEL_R2EFF and self.optimise_pre_run_r2eff):
if model == MODEL_R2EFF and self.exp_mc_sim_num != None:
self.interpreter.monte_carlo.setup(number=self.exp_mc_sim_num)
else:
r25000 - /branches/R1_fitting/auto_analyses/relax_disp.py