Dedicated code versus modelling environments: microsoft word is not a novel! You need to separate the model definitions from the coding, which has advantages in reproducibility, clarity, and collaboration. The emphasis, when creating modelling environments, is to ensure that the emphasis is on the biology. It follows that you should hide the complexity of the implementation from the users. These things increase the usability for the biologists.
Bob Murphy suggested then that biology is not a finite space. Additionally, you shouldn’t assume that the biologists understand the biology. The whole point of computational models is to create a level of understanding higher than the existing understanding. Finally, models predict observations and model building starts with observations. Putting a human between the observations and the model building isn’t necessarily the right thing to do, or that describing in words using an ontology is not necessarily the right thing to do. He thinks a traditional ontology is at best a stopgap that captures a snapshot of the knowledge at that point in time, and that we should think about non-word-based ontologies. Personal note: biological ontologies are, almost by definition, not snapshots: the state of our knowledge is always changing. If you don’t want just words, make use of the more complex first-order logic statements such as those available to people developing in OWL.Les Loew also made the important point that you shouldn’t take biologists out of the equation: we must keep the biologists as the focus of the meeting, as they are some of the most important groups ones we’re doing this for.
Back to James Glazier now. An ontology is not a model (in the sense of a computational biological model). However, it is a model of the domain you’re interested in. Neither is an ontology a syntax. An ontology is a logical structure that facilitates the model development and analysis. Their use case is for model sharing standards.
The virtual tissues discussed by Imran Shah can also be seen as multiscale modelling. His vision is a language for a specification for multiscale models. Want to start with CBO, but may not be limited to an ontologies. This should integrate easily with existing standards.
Glazier has a mock-up of a CBO-based ML, where each element name matches a CBO term. Personal Note: This could be problematic, for two main reasons: firstly, the example used labels rather than IDs (which was probably just for clarity), and secondly mock-ups of XML aren’t required, if the decision is instead to create instances of an ontology using RDF a la BioPAX, OBI or similar.
For CBO, Glazier is looking for something which provides agreed-upon hierarchical terms which we can then use to structure other applications/language. What cell behaviour ontologies exist now? GO, and Cell Physiology/Histology Ontologies, most of which are fairly fragmentary. There is a huge amount of behaviour in GO.GO stops fairly high up. We can fill in more specific terms under our remit.
Personal Note: We should be asking ourselves: Do we want to hang CBO under GO? If so, what about pathological terms? What about if CBO ends up more sematically rich than GO? Would it still be appropriate to align them even if the hierarchy looks different?
Please note that this post is merely my notes on the presentation. They are not guaranteed to be correct, and unless explicitly stated are not my opinions. They do not reflect the opinions of my employers. Any errors you can happily assume to be mine and no-one else’s. I’m happy to correct any errors you may spot – just let me know!