Variables

=================================================== Model set-up (see annotations in the driver script)

% Select pre-defined model

• Model_Example; Choose model example (1 = Pipeline example, 2 = PDGF model, 3 = CellNOpt example, 4 = Apoptosis model)

% Define optmisation options

• optRound; Number of optimisation round
• MaxFunEvals; Number of maximal function being evaluated (3000=default)
• MaxIter; Number of maximal iteration being evaluated (3000=default)
• Parallelisation; Use multiple cores for optimisation? (0=no, 1=yes)
• HLbound; Qualitative threshold between high and low inputs (0.5=default)
• Forced=1; Define whether single inputs and Boolean gates are forced to probability 1 (0=no, 1=yes)
• InitIC=2; Initialise parameters' distribution (1=uniform, 2=normal)

% Define plotting and saving (0=no, 1=yes)

• PlotFitEvolution; Graph of optimise fitting cost over iteration
• PlotFitSummary; Graph of state values at steady-state versus measurements (all in 1)
• PlotFitIndividual; Graph of state values at steady-state versus measurements (individual)
• PlotHeatmapCost; Heatmaps of optimal costs for each output for each condition absolute cost
• PlotStateSummary; Graph of only state values at steady-sate (all in 1)
• PlotStateEvolution; Graph of state values evolution over the course of the simulation (two graphs)
• PlotBiograph; Graph of network topology, nodes activities, and optimised parameters
• PlotAllBiographs; (Only for machines with strong GPUs) Plot all Biographs above

% Additional analyses after the optimisation with the default setting (0=no, 1=yes)

• Resampling_Analysis; Resampling of experimental data and re-optimise

• NDatasets; Number of artificial datasets from which to resample.

• LPSA_Analysis; Local parameter sensitivity analysis

• Fast_Option; Performing faster LPSA by stopping if fitting costs go over a set threshold value

• LPSA_Increments; Number of increments for LPSA. Increase for finer resolution

• KO_Analysis; Parameter knock-out analysis

• KONodes_Analysis; Node knock-out analysis

========================================================= Model & Optimization results (stored in Matlab workspace)

% Model information and inputs

• estim; Structure variable to store model information and results
• estim.Interactions; List of interactions in the model, 1st col = number of interaction, 2nd col = inputs, 3rd col = type of interaction (-> = activate; -| = inhibit), 4th col = outputs, 5th col = weight of interation (fixed value or variable to optimise), 6th col = Type of Boolean gate (N = no gate, A = AND gate, O = OR gate), 7th Col = Parameter range constraints (D = Default [0 to 1], L = Low [less than "HLbound"], H = High [higher than "HLbound"])
• estim.Input; List of input nodes (columns) for each experiment (rows)
• estim.Input_idx; List of indices of input nodes (columns) in the model for each experiment (rows)
• estim.Output; Experimental data for output nodes (columns) for each experimennt (rows); Note: NaN is used for missing data point(s)
• estim.Output_idx; List of indices of Output nodes (columns) in the model for each experiment (rows)
• estim.Output; The error of experimental data (e.g. SD or SEM) for output nodes (columns) for each experimennt (rows); Note: NaN is used for missing data point(s)
• estim.state_names; List of names for all nodes in the model
• estim.NrStates; Number of state/node in the model
• estim.NrParams; Number of optimising parameter in the model
• estim.param_index; Matrix of network information where 1st col = input indices, 2nd col = output indices, 3rd & 4th col = type of interactions (activate or inhibit, respectively), 5th col = type of Boolean gate (1 = AND, 2 = OR), 6th col = running number of Boolean gate in the model, 7th col = parameter range constraints (0 = default, -1 = low, 1 = high)
• estim.param_vector; Vector of optimising parameters
• estim.ma; Matrix of activation (read indicies in estim.state_names)
• estim.mi; Matrix of activation (read indicies in estim.state_names)
• estim.Aeq; Constrain equations for fmincon where Aeq*estim.param_vector = beq (constraints for sum of activating probabilities being 1)
• estim.beq; Constrain equations for fmincon where Aeq*estim.param_vector = beq (constraints for sum of activating probabilities being 1)
• estim.A; Constrain equations for fmincon where A*estim.param_vector = b (constraints for sum of inhibiting probabilities being less than 1)
• estim.b; Constrain equations for fmincon where A*estim.param_vector = b (constraints for sum of inhibiting probabilities being less than 1)
• estim.LB; List of lower bounds for parameters
• estim.UB; List of upper bounds for parameters
• estim.kInd: Indices of parameters
• estim.IdxInAct/estim.IdxOutAct/estim.IdxInInh/estim.IdxOutInh; Extracted vectors for Input/Output for activating/inhibiting reactions
• estim.BoolMax; Number of total Boolean gate(s) in the model
• estim.BoolIdx; Number of Boolean indies
• estim.BoolOuts; Indices of Boolean output node
• estim.FixBool; Indices of fixed Boolean variable
• estim.option; Default and customised optimisation options for fmincon
• estim.SSthresh; Threshold of fitting cost to accept the reach of steady-state

% Optimisation outputs

• estim.MaxTime; The maximum running time from the optimisation
• estim.AllofTheXs; State trajectory of each nodes during the optimisation (better representation in the plots)
• estim.MeanStateValueAll; Mean state value from multiple simulations
• estim.bestx; The best set of optimised parameter values

% Optimisation results as sub-structures (estim.Results)

• % Optimisation

• estim.Results.Optimisation.FittingCost; List all fitting costs

• estim.Results.Optimisation.FittingTime; List all optimisation time

• estim.Results.Optimisation.ParamNames; List all parameter names

• estim.Results.Optimisation.BestParams; List all best parameter values

• estim.Results.Optimisation.StateNames; List all state names

• % Fitting evolution

• estim.Results.FitEvol.PlotCosts; List all 3 re-run fitting costs

• estim.Results.FitEvol.Cost1/.Cost2/.Cost3; List fitting costs from the 3 re-runs

• % Resampling

• estim.Results.Resampling.Parameters; List all parameters

• estim.Results.Resampling.OptimisedParameters; List all optimised parameters with new re-sampled measurements

• estim.Results.Resampling.OptimisedSD; List the standard deviation from all optimised parameters with new re-sampled measurements

• estim.Results.Resampling.LargeSD; Determine if the SD are larger than the threshold

• estim.Results.Resampling.Costs; List all fitting cost during resampling process

• % LPSA (Local parameter sensitivity analysis)

• estim.Results.LPSA.ParamNames; List all parameter names

• estim.Results.LPSA.Identifiability; Vector determining whether each parameter is identifiable

• estim.Results.LPSA.LPSA_Increments; The number of parameter interval to estimate identifiability

• estim.Results.LPSA.p_SA; The list of parameters to perturb

• estim.Results.LPSA.cost_SA; The fitting cost after parameter perturbations

• estim.Results.LPSA.CutOff; The cut-off value of fitting cost to assess identifiability

• estim.Results.LPSA.Interpretation; Type of identifiability in estim.Results.LPSA.Identifiability ('1=Identifiable','2=Partially identifiable','3=Non-identifiable')

• % Knockout (interaction)

• estim.Results.KnockOut.Parameters; List of parameters to knock-out parameters

• estim.Results.KnockOut.AIC_values; List of AIC_values after parameter knockout

• estim.Results.KnockOut.KO_effect; List of interpretation if knockout has a substantial effect

• estim.Results.KnockOut.Interpretation; Interpreter for knockout results ('0 = no KO effect','1 = KO effect')

• % KnockoutNode

• estim.Results.KnockOutNode.Parameters; List of nodes to knock-out

• estim.Results.KnockOutNode.AIC_values; List of AIC_values after node knockout

• estim.Results.KnockOutNode.KO_effect; List of interpretation if knockout has a substantial effect

• estim.Results.KnockOutNode.Interpretation; Interpreter for knockout results ('0 = no KO effect','1 = KO effect')

===========================================================