Piecewise parameter estimation in COPASI

Examples for piecewise parameter fitting with COPASI

This page contains several examples files that allow to ascertain how a new feature in COPASI can be used to set up piecewise parameter fits that can be used for both, stochastic and deterministic models. Each example will be given as archive, that archive contains the COPASI as well as all the experimental data associated with it. To run the examples simply:

  • download the archive
  • extract the archive (including the subfolder contained in it)
  • change into the subfolder and open the COPASI file (*.cps)
  • within COPASI click on Tools\Create Events For Timeseries Experiment. Once this is done COPASI will turn all the individual rows of the timeseries experiment file into events that are added to the model.
  • select the parameter estimation task, and press run. Learn more
  • the results can be examined using plots or the result widget. Learn more

Examples for parameter estimations of stochastic models

These examples have been described originally in Zimmer and Sahle (2012). They can be reproduced using the method outlined above.

For each example the archive will contain two models: the first one being the base model <file>-model.cps, it does not yet contain any events. The second model <file>-model1.cps contains the events added from zr1.txt. To select a different set of measurements we deleted events from model1.cps, replaced the experiment set of zr1.txt with the one of the other files and then regenerated the events as described above.

For each of the files below we give a visualization of the first to time series to give you an idea for the data

Immigration-Death

Example of the first two timeseries

In this archive <file> will be called id so for example id-model.cps.

Download »

Lotka-Volterra - full observation

Example of the first two timeseries

In this archive <file> will be called lv so for example lv-model.cps.

Download »

Lotka-Volterra - partial observation

Example of the first two timeseries

In this archive <file> will be called lvp so for example lvp-model.cps.

Download »

Examples for piecewise parameter estimations of deterministic models

These examples allow to compare how the fitness landsape differes when simply applying parameter estimation (thus using a least square objective function), or using the method outlined above (thus using the MSS objective function). To compare the differences you would run the parameter estimation, once with the events being added (i.e. MSS) and then once again with the events being deleted (i.e. LSQ).

One dimensional plots of the fitness landscape can be obtained, by first finding a good fit, then changing the optimization method to Current Solution Statistic (which just evaluates the objective function), and then defining a Parameter Scan. This has already been done in the models contained in the archives below.

Lotka-Volterra

This example contains four models: lv.cps,lvBIG.cps, lvmss.cps, lvmssBIG.cps. The files ending in BIG use a bigger parameter range.

Download »

Fitz-Hugh Nagumo

This example contains four models: fhn.cps,fhnBIG.cps, fhnmss.cps, fhnmssBIG.cps. The files ending in BIG use a bigger parameter range.

Download »

Calcium

This example contains seven models: calcium.cps,calciumMSS.cps, calciumMSS_2steps.cps, calciumMSS_4steps.cps, calciumMSS_10steps.cps, calciumMSS_20steps.cps, calciumMSS_50steps.cps. The different step files denote how many events have been included.

Download »