Re: r25000 - /branches/R1_fitting/auto_analyses/relax_disp.py -- August 20, 2014 - 18:20

Hi Edward.

I partly implemented this.

But I think some of the functionality still missing in the run(self)
part of the analysis.

I think this part will give problems.

--------
            # Select the model.
            self.interpreter.relax_disp.select_model(model)

            # Copy the R2eff values from the R2eff model data pipe.
            if model != MODEL_R2EFF and MODEL_R2EFF in self.models:

self.interpreter.value.copy(pipe_from=self.name_pipe(MODEL_R2EFF),
pipe_to=model_pipe, param='r2eff')

            # Calculate the R2eff values for the fixed relaxation time
period data types.
            if model == MODEL_R2EFF and not has_exponential_exp_type():
                self.interpreter.minimise.calculate()
---------

Just reading in R2eff values by relax_disp.r2eff_read, will not give
enough information?

best
Troels

2014-08-18 18:46 GMT+02:00 Edward d'Auvergne <edward@xxxxxxxxxxxxx>:
> I would suggest a slightly different solution.  If the data pipe sent
> into the auto-analysis already contains R2eff values and errors, then
> skip that model optimisation.  The flag can be used anyway to force
> the R2eff parameter optimisation.
>
> On another note, this flag is not necessarily related to the pre_run
> and can be used without a pre-run.  For example, a user like Nikolai
> could have files of R2eff values that are read in via the
> relax_disp.r2eff_read.  So the flag would be better called
> optimise_r2eff.  You currently use the flag when no pre-run data is
> used, so that part needs no changing.
>
> Cheers,
>
> Edward
>
>
>
> On 12 August 2014 15:23,  <tlinnet@xxxxxxxxxxxxx> wrote:
> > 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:
> >
> >
> > _______________________________________________
> > relax (http://www.nmr-relax.com)
> >
> > This is the relax-commits mailing list
> > relax-commits@xxxxxxx
> >
> > To unsubscribe from this list, get a password
> > reminder, or change your subscription options,
> > visit the list information page at
> > https://mail.gna.org/listinfo/relax-commits
>
> _______________________________________________
> relax (http://www.nmr-relax.com)
>
> This is the relax-devel mailing list
> relax-devel@xxxxxxx
>
> To unsubscribe from this list, get a password
> reminder, or change your subscription options,
> visit the list information page at
> https://mail.gna.org/listinfo/relax-devel