Model creation

  1. Either via graphical model designer or a text editor
  2. Chemical reaction based (ODE creation done by the PottersWheel chemical compiler)
  3. ODE based (optionally with reaction network reconstruction)
  4. Import from SBML model or a regular custom structure
  5. Rule based modeling to cope with combinatorial complexity
  6. Algebraic equations (assignment rules, start value assignments)
  7. Support of model families without redundant modeling

Model integration

  1. The differential equations of a model are dynamically compiled as C files
  2. 14 integrators supported (6 FORTRAN, 7 Matlab, 1 C)
  3. Very fast model integration allows for a real time modeling and fitting experience

Model investigation

  1. Equalizer allows to change parameter values via sliders in real time
  2. Input Designer allows to change the characteristics of external driving functions, like a continuous, pulsed, or ramp stimulation
  3. Automatic investigations like sensitivity analyses including control coefficients

External driving input

  1. Strong support of external input functions which drive the dynamical system
  2. Input is either given analytically or based on experimental data

Experimental data

  1. Incorporate experimental data stored in xls or txt files or an SQL database
  2. Automatic detection of data column names and mapping dialog for model observations
  3. Automatic use of already mapped data sets
  4. Support of multiple stimulations within one experiment
  5. Dose- or stimulus-dependent view possible
  6. Estimation of experimental standard deviation
  7. Outliers dialog to remove single data points after visual inspection

Fitting

  1. Support of powerful deterministic and stochastic optimization algorithms
  2. Multi-Experiment fitting: One model is fitted simultaneously to several data sets. Parameters with a unique or data set depending meaning are distinguished.
  3. Several fit sequences available, like n fits each starting from the last fit with disturbed parameter values
  4. Fit sequence starting from locations within the complete parameter space
  5. Qualitative or soft constraints can be specified
  6. Fit sequences can be run on a cluster using the parallel computing toolbox

Fit sequence analysis

  1. Best fit selection
  2. Histograms and box plots of fitted parameter values
  3. Linear correlation analysis including principal component analysis (PCA) in order to detect linear non-identifiabilities (see Plug Ins section for detection of non-linear non-identifiabilities)
  4. Hierarchical clustering of the parameter space
  5. Analyis of custom derived parameters, which are functions of parameters
  6. ScatterMan: Manual investigation of pairs and triples of fitted parameter values

Single fit analysis

  1. Detailed residual analysis
  2. Statistical tests for chi square values

Reporting

  1. Each analysis or fit can be appended as a section to a report
  2. Optional with graphical model visualization and a list of reactions
  3. The order of sections can be changed at any time
  4. All figures of a section can separately be investigated
  5. Support of PDF-Latex (recommended), MS Word, and HTML reports

Sessions

  1. The complete current working state can be saved and reloaded any time
  2. Exchange of sessions between researches to exactly reproduce modeling efforts
  3. A modeling session comprises an arbitray set of repository models and data sets and currently combined model-data couples for fitting

Application programming interface and macros

  1. Rich set of Matlab functions to use PottersWheel functionalities within custom Matlab programs
  2. Macros support an automated and documented way to model efficiently
  3. Comprehensive, uptodate documentation via 'help FunctionName' or online

Graphical user interfaces

  1. Modeling requires no Matlab knowledge due to many user friendly graphical interfaces
  2. Steep learning curve
  3. Modeler has more time to focus on modeling than on technical details