Skip to content

Applying the Benchmark Models using Data2Dynamics

Jump to bottom
Clemens Kreutz edited this page Oct 25, 2018 · 5 revisions

Data2Dynamics (D2D) is a Matlab-based modelling framework tailored to ODE-based modelling of cellular processes. D2D offers comprehensive functionality and is available at https://github.com/Data2Dynamics/d2d.

Data2Dynamics provides an comprehensive set of application and toy models, including the presented benchmark collection, which are stored in the folder "Examples". This comprises model and data definition in the D2D format.

In the following, we describe important tasks for usage of the benchmark collection in D2D:

Setting up D2D

The following steps are required for setting up D2D

  1. Download of Data2Dynamics, e.g. from https://github.com/Data2Dynamics/d2d/archive/master.zip
  2. Adding the D2D basic folder e.g. via addpath('D2D_LOCATION/d2d/arFramework3')
  3. Initialization of D2D via arInit

Since D2D generates C-code which is compiled for usage in Matlab, the so-called mex compiler has to be properly configured in Matlab.

Check the D2D wiki for further details about installation of Data2Dynamics.

Working with a specific benchmark model

The most straight forward way of using a benchmark models is switching the Matlab work folder to a specific benchmark model e.g. via cd('D2D_LOCATION/d2d/arFramework3/Examples/Becker_Science2010').

D2D provides a brief file termed Setup.m containing the commands for compiling and loading a model. Execution of this commands via Setup is sufficient for loading the model. Then basic D2D command like plotting (via arPlot) or fitting (via arFit) can be applied.

Working with all benchmark models

D2D provides functionality for selecting the benchmark collection subset and for comprehensive compiling of the model files. We suggest the following procedure:

  1. arCopyBenchmarkModels copies all 20 model folders to the current working directory
  2. arCompileAllSetups can be used to start compiling all models

Compiling the model files in many cases only takes some seconds. However, there are some models where compilation is time-consuming because a huge number of symbolic operations are necessary. Because compiling is required only once, we therefore recommend to compilation of all models over night, e.g. via arCompileAllSetups. The required time for the individual models depend on the computer and on the operating system. The following models are computationally demanding:

  • Bachmann: 2-3 minutes
  • Chen: 10 hours
  • Crauste: 2-3 minutes
  • Fiedler: 15 minutes
  • Fujita: 15 minutes
  • Isensee: 30 minutes
  • Lucarelli: 15 minutes
  • Merkle: 30 minutes
  • Reelin: 1-2 minutes
  • Schwen: 1-2 minutes
  • Sobotta: 40 minutes

Simulated data

If simulated data is intended to be used for benchmarking purpose instead of the experimental data it can be generated via arSimuMeasurements. This function generates data points at the same experimental conditions with normally distributed errors with the same standard deviation as specified in the model or data.

Clone this wiki locally