> The following is rather speculative - perhaps there is a way to do this > already that I'm unaware of. I'm just fishing for comments.
always a good thing to be doing :-) > If you have a model which contains only math, and say you want to use it > as a component. It has no parameters, and you need to plug in a whole > plethora of initial conditions etc. Is there any way to avoid doing this > manually? If you had a file containing all these parameters and values > and the variables/components that they pertain to, it would be good to > be able to simply import these into the model automatically. I'm not aware of any tools that create these connections automatically for you, either from imported components or local components within a model. Perhaps this is something Andrew has thought about with PCEnv, currently the only GUI tool that supports model imports. If you are into scripting then it would probably be fairly easy to write a script which add connections for variables with matching names or something similar. > The most important use for this parameter importing feature could be in > experimentation with the models. If you had a set of parameters that you > wanted to change over a range (for example, ligand concentration in a > GPCR signalling system,) and take the results that the model outputs > (you might want to know cAMP production rates by adenylate cyclase, PKA > activation levels, desensitization rates, etc. etc.) you could run this > 'experiment file,' and get the results output to CSV. It would probably > take an awfully long time to integrate the model several hundred times > if you wanted that many data points, but at present, you'd have to > manually change the values you wanted to change, integrate, export to > CSV, then do it over again. There are existing tools that let you do this in an automated fashion - either something like sensitivity analysis and/or parameter optimisation. Tools like JSim and Virtual Cell do this for both (although possibly you need the latest version of VCell to get the sensitivity analysis). When you use such tools to perform the types of experimentation you are talking about you don't really need to modify the underlying CellML model, you simply identify which variables you want to vary and the range over which to vary them. The trick comes in terms of model curation and the idea of a reference description - if you need a reference description for each parameter value version of the model then you would need to somehow encapsulate all the variable ranges in CellML. I'm not sure when that would ever be necessary, but maybe there are some smart ways to arrange all these models without having to repeat too much information... Andre. _______________________________________________ cellml-discussion mailing list [email protected] http://www.cellml.org/mailman/listinfo/cellml-discussion
