Is it possible to develop a standardized format for expressing and sharing Cell Biological models?

Discussion Leaders from Workshop 3

Nega Bagheri

Holly Goodson

Jian Liu



Existing standards to be considered in this context:

SBML - Systems Biology Markup Language

vCell - the Virtual Cell

PySB - systems biology in python

CellML - from physiome project for cardiac models with biochemical components and true multiscale components


Issues and challenges

Models are subjective, context-dependent, personal, etc… but so are gene ontology databases

Do not need a database or standard unless someone wants to reuse a model and no one wants to reuse a model

We don’t trust another person’s model

More work to understand someone else’s model than to just make your own model

Competitive landscape does not favor easy model sharing

Different models may be devised for the same system because they are asking different questions and therefore modeling different aspects of the system

Different types of data call for different models to understand them

Confusion between standardizing the model and standardizing the simulation or solution method used to make predictions from the model

Perception among many participants that SBML has involved a huge cost by developers but has had minimal impact on actual biological researchers

SBML is missing important biological concepts such as polymers or organelles.   Other standardized formats may share this limitation.

Concerns about maintainability - how resuable will code be if libraries are no longer available or are replaced by other standards?


What we need

Need to develop incentive: community investment; journal/editors to require code to be published; NSF/NIH to impose expectations

What basic principles should a model be able to reproduce?

Need a control for a model

A way to formalize information in papers to be used automatically in simulations independent of the model – having a clear way to describe data would make model comparison easier

Instead of formalizing models, maybe what we really need is a way to standardize reporting of findings in papers

A repository for models that would give credit to the originator when the model is shared

Ways to describe a system in a machine readable way that is simulation-agnostic.

Ways to separate model specification from implementation details

Community involvement in database design