| Popis: |
Increasingly, the biological modelling community is looking for ways of handling complex and possibly dynamic structures in their models. On the one hand, the pathway-modelling groups (SBML and CellML) seek to move up levels of organisation, while on the other hand the "Virtual X" communities (where X is some organism or organ) need to represent the organisational structure within their object of study. To date, various approaches have been proposed, such as the SBML Level 3 'comp', 'spatial', 'array' and 'dyn' extension packages, and domain-specific elements in languages such as NeuroML, but there is no generic approach intended to be adopted by the various communities. We have designed and implemented such a generic approach in the Simile modelling environment ("http://www.simulistics.com":http://www.simulistics.com). Simile combines a diagrammatic notation for ODEs (in terms of stocks, flows and subsidiary equations), combined with a UML-like notation for representing classes of object and associations between them. No domain-specific concepts are built in, so the object classes can represent anything from molecules to planets, and the associations are specified by the modeller. For example, Simile has no spatial concepts built in, but the modeller can create a class called (perhaps) 'spatial unit', and specify a regular (2- or 4-neighbour) square grid, a 3D array, a hexagonal grid or arbitrary polygons, simply by changing the expression which defines when the association exists between any pair of spatial units. Containership is represented simply by drawing one class inside another (so you can easily show that molecules are inside cells, and cells are inside organs), and objects can be dynamically created and destroyed during the simulation. Because Simile generates compiled C++ code for simulating model behaviour, it is capable of handling very complex models (1 million+ objects). We thus consider that Simile demonstrates the feasibility of a generic approach for handling disaggregation in biological models. We suggest that the widespread adoption of such an approach across the various biological modelling communities would greatly improve the efficiency and effectiveness of future model development as the models become increasingly complex. |
Full text from SpringerLink