cmb_options = cmb_default_options(flag_artificial)
flag_artificial (Boolean, default 0) options for model with artificial data
Returns matlab struct with default options for model balancing
Fields and default values are shown below
Note that this function reads information about priors from the prior table (see 'cmb_prior_file')
and stores them in cmb_options.quantities; there they can be modified by the user
Algorithm settings
.run = 'my_run'; % name of estimation scenario (freely chosen by the user)
.kinetic_data_set = 'original'; % name of kinetic data set (freely chosen by the user)
.ecm_score = 'emc4cm'; % ECM score function, see 'help ecm_scores'
.initial_values_variant = 'average_sample'; % strategy for choosing initial values (see below and 'help cmb_estimation')
possible choices: 'average_sample', 'preposterior_mode',
'random', 'true_values', 'given_values'
.enzyme_score_type = 'interpolated'; % 'quadratic' (alpha=1), 'monotonic' (alpha =0), 'interpolated'
(only "monotonic" guarantees that MB is convex!)
.enzyme_score_alpha = 0.5; % stringency parameter alpha for enzyme_score_type='interpolated'
.beta_ln_c_over_km = 0 % beta = 1/sigma for penalty on sum_til ln(c_i(t)/km_il) = 1/ln(geom std of c/KM)!!
.parameterisation = 'Keq_KV_KM_KA_KI'; % options: 'Keq_KV_KM_KA_KI', 'KV_KM_KA_KI';
.use_kinetic_data = 'all'; % 'all', 'only_Keq', 'none'
.score = 'neg_log_posterior'; % options: 'neg_log_posterior', 'log_neg_log_posterior';
.use_gradient = 0; % flag: set opt.SpecifyObjectiveGradient = true in fmincon optimisation
.use_safe_optimisation = 1; % flag: run several rounds of optimisation (until convergence)
.random_seed = 1; % random seed for optimisation
Display and output options
.verbose = 0; % flag: verbose
.optim_display = 'off'; % flag: show graphical output during optimisation
.display = 1; % flag: display table with scores after each optimisation round
.show_graphics = 1; % flag: show graphics
.plot_true_vs_data = 0; % flag: show extra plots: true values against data values
(only in the case of artificial data)
.save_results = 1; % flag: save results to files
.save_graphics = 1; % flag: save graphics to files
Options for generating artificial data (see 'help cmb_model_artificial_data')
.use_artificial_noise = 0; % flag: generate artificial state data with noise
.use_kinetic_data_noise = 1; % flag: generate artificial kinetic data with noise
.prior_variant = 'original_prior'; % prior variants (see cmb_model_artificial_data)
% also 'broad_prior', 'broad_prior_around_zero',
% 'prior_around_true_values', 'broad_prior_around_true_values'
Bounds and distributions for model variables
Default values are taken from the prior file, see 'cmb_prior_file'
.use_bounds Boolean (default 1) -> if set to 0, lower and upper bounds are ignored for all variables
.use_pseudo_values Boolean (default 0) -> if set to 0, no pseudo values are used
.use_crosscovariances Boolean (default 0) -> if set to 0, crosscovariances (between types of kinetic constants) are ignored
.quantities.Keq.max new value for upper_bound(ind.Keq);
.quantities.Keq.mean_ln new value for prior_median_ln(ind.Keq);
.quantities.Keq.std_ln new value for prior_geostd_ln(ind.Keq);
.quantities.KV.min new value for lower_bound(ind.KV);
.quantities.KV.max new value for upper_bound(ind.KV);
.quantities.KV.mean_ln new value for prior_median_ln(ind.KV);
.quantities.KV.std_ln new value for prior_std_ln(ind.KV), default log(5);
.quantities.KM.min new value for lower_bound(ind.KM);
.quantities.KM.max new value for upper_bound(ind.KM);
.quantities.KM.mean_ln new value for prior_median_ln(ind.KM);
.quantities.KM.std_ln new value for prior_geostd_ln(ind.KM), default log(5 mM)
.quantities.KA.min new value for lower_bound(ind.KA);
.quantities.KA.max new value for upper_bound(ind.A);
.quantities.KA.mean_ln new value for prior_median_ln(ind.KA);
.quantities.KA.std_ln new value for prior_geostd_ln(ind.KA);
.quantities.KI.min new value for lower_bound(ind.KI);
.quantities.KI.max new value for upper_bound(ind.KI);
.quantities.KI.mean_ln new value for prior_median_ln(ind.KI);
.quantities.KI.std_ln new value for prior_geostd_ln(ind.KI);
.quantities.Kcatf.min new value for lower_bound(ind.Kcatf);
.quantities.Kcatf.max new value for upper_bound(ind.Kcatf);
.quantities.Kcatf.mean_ln new value for prior_median_ln(ind.Kcatf);
.quantities.Kcatf.std_ln new value for prior_geostd_ln(ind.Kcatf);
.quantities.Kcatr.min new value for lower_bound(ind.Kcatr);
.quantities.Kcatr.max new value for upper_bound(ind.Kcatr);
.quantities.Kcatr.mean_ln new value for prior_median_ln(ind.Kcatr);
.quantities.Kcatr.std_ln new value for prior_geostd_ln(ind.Kcatr);
.quantities.v.max new value for upper_bound(v), default 100 mM/s
.quantities.c.min new value for lower_bound(ind.c);
.quantities.c.max new value for upper_bound(ind.c);
.quantities.e.max new value for upper_bound(ind.u);
.quantities.Aforward.min new value for lower_bound(ind.A), default 0.0001 kJ/Mol
.quantities.Aforward.max new value for upper_bound(ind.A);
Geometric standard deviations for priors
(also used for sampling "true" values, in the case of artificial data)
default values are taken from the prior file, see 'cmb_prior_file'
.metabolic_prior_c_geom_mean
.metabolic_prior_e_geom_mean
.metabolic_prior_c_geom_std
.metabolic_prior_e_geom_std
Geometric standard deviations for data values (describing measurement uncertainties)
- used in 'state_data_to_data', to complete standard deviations not given in the data files
- also used in 'cmb_generate_artificial_data', for generating noise in artificial data
.data_C_geom_std default from prior table
.data_E_geom_std default from prior table
.data_V_geom_std default 1.2 % currently not used
.data_kin_geom_std default 1.5
Distributions for "true values" inartificial data (in 'cmb_generate_artificial_data')
.metabolic_artificial_c_geom_std default 1.5;
.metabolic_artificial_e_geom_std default 1.5;
also used in 'cmb_generate_artificial_data':
.quantities.mu0.std default RT * log(5);
.quantities.KV.mean_ln
.quantities.KV.std_ln
.quantities.KV.min
.quantities.KV.max
.quantities.KM.mean_ln
.quantities.KM.std_ln
.quantities.KM.min
.quantities.KM.max
Possible initial value variants (option 'cmb_options.initial_values_variant'):
'polytope center' (default) center of mass of some random LP solutions
'preposterior_mode' use preposterior mode for kinetic constants and metabolite levels
'flat_objective' pick a point in the solution space by using matlab's fmincon with a uniform objective function
'random' initialise parameter vector with random values
'true_values' true values: only with artificial data
'given_values' use existing cmb_options.init
'average_sample' first run a model balancing with concentrations and fluxes averaged over all samples;
and using initial value option "preposterior_mode"; use result to initialise values