This script is dedicated to allow one to quickly and efficiently run a Copasi parameter estimation task on a computing cluster. Running 1000 parameter estimations in minutes and automatically summarising the results allows you to test a lot of hypothesis in a single day! Please enjoy and fork/pull request if you can see some obvious improvements!
- LSF or SGE cluster
- COPASI, perl and Bash installed on the cluster.
- R itself and gplots and ggplot2 packages installed in it.
- Prepare your COPASI model file (see example in the model folder):
- Setup Parameter Estimation task with the default Report
- The Parameter Estimation Report file must be named param-est-report.txt
-
Copy your .cps file in the model directory (don't forget to remove all example files from it) together with experimental data file
-
From root directory run:
sh coparaest.sh n c
where n is number of parameter estimations and c is your cluster queuing system (sge or lsf). This will start n parameter-estimation jobs, one get-obj-values job and one analyse-results job.
(after analyse-results job is finished):
-
out, err - these folders contain cluster output files with its output and possible errors
-
results/obj-values.txt - this file contains indecies of parameter estimations (first column) sorted by their objective values (second column) with the best estimation in the first row.
-
results/ind/estd-params.txt - this file contains all estimated parameters for parameter estimation with ind index.
-
results/model-correlations.pdf - heatmap of correlations between models in the top 10 estimated parameter sets
-
results/param-correlations.pdf - heatmap of correlations between parameters in the top 10 estimated parameter sets
-
results/param-correlations.pdf - variance of parameters in the top 10 estimated parameter sets
Authors: Vladimir Kiselev, Marija Jankovic
Acknowledgments: Martina Fröhlich, Nicolas Rodriguez