Small scale project, 2021 Systems biology simulation
Scenario
Researcher
Lilija Aprupe-Wehling, PhD student in Modeling of Biological Processes, Centre for Organismal Studies
Initial Problem
- Parameter fitting of a spatial biochemical reaction model using a particle swarm algorithm suffers from two bottlenecks:
- Changing the model, parameters or data used in a fit requires tedious and error prone changes to a Python script
- Running the algorithm on a single cpu core can take too long to converge to a good solution
Outcome
What we did
- Wrapped the functionality in a user-friendly Python module: sme_contrib.optimize
- Made the fitting thread-safe, then enabled multi-threaded fitting using Python multiprocessing
- Set up a github repository with continuous integration for automated testing and deployment of the code to PyPI
- Provided access to a HPC node for testing the module and evaluating the multi-threaded performance
Results
- Faster and simpler process for the user to set up parameter fitting
- Faster results thanks to the multi-threaded implementation
- This fitting functionality can easily be used and extended by other researchers