Skip to main content

A python module for performing identifiability analysis using copasi

Project description

Installation

Use:
pip install pydentify
at command prompt or download and use:
>>>python setup.py install
from the downloaded directory


Author: Ciaran Welsh. Email: c.welsh2@newcastle.ac.uk

Pydentify is a python package intended for use by systems biologists that
are looking to fit experimental data to their models. Because of the dimensionality of the
models in question the optimization process is diffult. Often parameters
cannot be uniquely defined by the optimization problem. This is called a non-identifiability.
COPASI (Mendes et al 2009) is a software package that enables one to easily perform parameter estimation.
Pydentify uses COPASI as a 'parameter estimation engine' in order to calculate the profile likelihood
method of identifiability analysis (Schaber, Biosystems 20012).


Instructions


Before using this script to perform Identifiability analysis
in Copasi, as described by Jorg Schabber, you must first define an
appropriate CopasiML using the GUI.

Ensure you have done the following:
1) Ensure the parent .cps file is located in the SAME directory as the data files used for parameterization and that no other text files are present in this directory. Its best to have a dedicated folder for your identifiability analysis.
2) Perform parameter estimation to locate the global minimum and update model. Alternatively you can use the repeat feature in COPASI's scan task to perform multiple parameter estimations and have the results written to a report. The latter procedure is best accomplished using a computer cluster.
3) Create a new report from COPASI's 'output specifications' window that contains any, but only one estimated parameter and the RSS value. The latter can be found by checking the 'expert mode' button and then going to: 'ModelList>Root>TaskList>ParameterEstimation>ParameterEstimation>BestValue'.
4) Ensure you are using Hook and Jeeves for all secondary parameter estimations.
5) Create a scan in the parameter scan subtask with any estimated parameter:
6) Change the scan subtask to parameter estimation
7) Check the 'Log' checkbox to scan on a log scale
8) Check the executable box in the top right hand corner of the scan window
9) Define a report using the 'ProfileLikelihood' report that was previously defined. Name it anything and uncheck the 'append' and 'confirm overwrite' buttons.
10) Delete any other reports that you have defined in other subtasks
11) Delete any parameter sets or events that you have defined
12) Ensure the time, volume and quantity units are properly defined as they are used in some calculations with Avogadro's constant.

Then use:

PL=ProfileLikelihood(<copasi_file>)
PL.run(mode='slow')

<copasi_file> : absolute path to your configured copasi file
or:
MLP=MultiProfileLikelihoods(<copasi_file>,[index],results_dir=<results_directory>)

[index] : python list for index of parameter estimation runs you want to calcualte profile likelihoods for
results_dir: absolute path to a directory containing parameter estimation output from copasi
results_file: absolute path to file containing patameter estimation data
Which depends on the module 'InsertCopasiParameters' provided with pydentify.py.

#Note that for some reason this program will not run from a Dropbox (or similar) directory

#Also, if you have to stop running the script half way through and attempt to restart,

To visualize results:

If using pydentify.ProfileLikelihood():

P=pydentify.Plot(copasi_file,RSS_i)
copasi_file: an appropiately configured copasi file
RSS_i: the RSS value for the original parameter estimation
P.plot_all(savefig=True)

If using pydentify.MultiProfileLikelihood():

MP=pydentify.MultiPlot(copasi_file,results_dir=<results_directory>)

OR

MP=pydentify.MultiPlot(copasi_file, results_file=<results_file>)

where:
copasi_file: an appropiately configured copasi file
results_dir: full path to a results directory (containing multi files of parmaeter estimation data in txt format, such as the output from copasi parameter estimation)
results_file: full path to a results file containing parameter estimation data (xlsx/xls/csv/txt)

Then
MP.plot_index(0,savefig=True) #to plot index 0 of your multi profile likelihood
or
MP.plot_all_indexes(savefig=True,multiplot=False) ' #to plot all indexes
MP.plot_all_indexes(savefig=True,multiplot=True) #to plot all profiles on the same graph for comparison. Resutls are viewed in the largest index folder as this is an iterative process

References
Mendes, P., Hoops, S., Sahle, S., Gauges, R., Dada, J. and Kummer, U. (2009) 'Computational modeling of biochemical networks using COPASI', Systems Biology, pp. 17-59.
Schaber, J. (2012) 'Easy parameter identifiability analysis with COPASI', Biosystems, 110(3), pp. 183-185.

Project details


Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page