Hi all,

I have had a look through the proposals at 
http://sbml.org/wiki/SBML_Level_3_Efforts . I don't know if there is a 
draft for the 'core' of SBML level three yet, although it seems the plan 
is that it will not change much from level 2.

I think that under the current approach of splitting metadata and data 
in CellML such that anything which alters the mathematical 
interpretation of the model is data, and anything else is metadata, 
some, but not all, of the extensions for SBML 3 would be data in CellML.

Going through each extension:
* Diagram Representation: This appears to be pure metadata information.
* Spatial Features: This also appears to be pure metadata information.
* Hierarchical Model Composition: This is similar in intent to the 
import functionality already present in CellML.
* Multicomponent species: This seems to address a number of related 
problems regarding the representations of reactions which act on more 
than one species, where the species differ only by changes in 
phosphorylation state or location, or some other property like this. 
This is probably too domain specific for CellML to deal with in this way.
* Dynamic structures: This aims to allow for the structure of models to 
change dynamically (e.g. to create multicellular models). The module 
documentation notes that "arrays and sets are alternative proposals for 
roughly the same kind of capability, and it is likely that only one will 
ultimately be chosen as a supported SBML Level 3 language extension".
* Arrays / Sets: This extension aims to describe something fairly 
similar to the goals of the CellML proposals to add additional data 
types other than real numbers.
* Parameter Sets: This seems to be a way to split the model mathematics 
away from parameter definitions. This is commonly already done in CellML 
using imports.
* Alternative Reactions: This aims at being able to specify what type of 
model is being expressed. It has some similarities in intent to the 
simulation metadata, but it does actually alter the interpretation of 
the model. I think that this only makes sense in SBML because of the 
domain specific representation in SBML can be mapped to multiple 
different non-domain specific mappings. In CellML, this mapping has 
already been performed. However, the extension in SBML does imply the 
ability to represent other types of models, such as stochastic models.
* Distributions: The representation of statistical distributions is 
something which CellML does not currently provide any explicit provision 
for. However, it may be something that could be fitted into CellML 
without any major changes, so is something worth studying. This may be 
complicated by the operators present as predifined symbols content 
MathML - we may need to define some new semantics to allow for this type 
of model.

Best regards,
Andrew

_______________________________________________
cellml-discussion mailing list
cellml-discussion@cellml.org
http://www.cellml.org/mailman/listinfo/cellml-discussion

Reply via email to