Date(s) - Tuesday, April 2nd, 2019
7:00 pm - 8:15 pm
For many scientists the mathematical modelling process is surrounded by magic; given some data and a magical wand, a model appears that explains the process. And sure, there is some room for creativity when setting boundaries, defining state variables, and deriving rate equations for a model, specifically when no clear mechanism for the process is available, but there is no excuse for not making the modelling process explicit. The modelling workflow should be transparent and reproducible.
One can split up the modelling process in three parts: model construction, model validation, and model dissemination/publication. Model construction and validation are typically linked to experimental data, and it should be clear what data is used where in the process. For this, all data used for model construction should be made available and explicitly labelled, similarly for the validation data set. Ideally the model parameterisation should be reproducible, given the data and model equations. An independent validation experiment should be used for model validation, and the data set should be compared with the model prediction.
Model construction and validation can run over multiple iterations between experiment and model, and a good versioning system is important. In addition, the number of data and model files, and their links, quickly becomes large and a good data management system is important to keep track of the process.
The FAIRDOMHub (https://fairdomhub.org) is a data and model management platform that has been developed for over a decade in large systems biology programmes. The platform makes it easy to show the links between data and model files, in so-called ISA structures (investigation-study-assay), and it is straightforward to make the workflow transparent, and reproducible. In addition, models can be simulated from within the platform in a web browser (using the integrated JWS Online system), and model simulation experiment descriptions (SED-ML) scripts can be generated on the fly (https://jjj.bio.vu.nl/models/experiments/).
In the webinar I will show examples of modelling workflows, starting from experimental data and model construction, up to model publication. I will illustrate how using the FAIRDOMHub and JWS Online makes it easier to stick to standards required by scientific journals, making the dissemination/publication a lot simpler, and keeping the process transparent and reproducible.
For more information please visit http://reproduciblebiomodels.org/seminar
The Webinar can be accessed on ZOOM via the link: https://washington.zoom.us/j/992899771
Webinar is at 7PM UTC – (8PM BST, 9PM CEST)