Toward Systems Biology
May 30 - 31, June 1, 2011
Grenoble
Energy as Syntax
There is a growing interest in domain-specific modeling/programming
languages in the context of systems and synthetic biology - and in the
larger view, in formal approaches to complex systems, where syntax
meets dynamics (and learning). We have developed the concepts of
rule-based modeling of biomolecular networks and the accompanying
kappa language. The approach is gaining traction in the systems and
synthetic biology communities. Here is a Nov 2009 Nature feature
article mentioning our approach. Here is another one (Jul 2009). Yet
another one in a recent issue of Nature Methods (Feb 2011) - with a
vibrant endorsement of our methods! In the realm of synthetic biology,
the Edinburgh IGEM 2010 team has won the "best model" prize for a
comprehensive rule-based modeling effort.
This increasing recognition is part of a broader realization that in
order to address decentralized dynamics of high complexity and
connectedness, one has to go beyond basic descriptive tools such as
Markov chains and differential equations. New modeling situations
present a diversity of structures and scales where the
representational challenge can be insurmountable without abstract and
structured syntaxes to describe the dynamics of interest. Part of the
modeling activity gradually morphs into a sort of domain-specific
programming. Now, enriching the representational apparatus of a
modeling domain is not just a way to make the modeling more agile and
far-reaching. In a way that has been long recognised in the context of
programming, one can lean on syntactic structures to develop various
analyses that would otherwise be unfeasible.
With this in mind, and in order to advance the specific modeling and
analysis of biomolecular networks, I will explain how now we are
trying to borrow some structuring features from biophysics and develop
a modeling language where energetic and thermodynamic constraints are
put front and center. The idea is to develop energy as a syntax. That
is to say, to investigate how one can structure and program the
dynamics of rule-based stochastic binding systems by the means of
local energy functionals describing their equilibrium properties. In
this new fashion of modeling, the dynamics is inferred from the
statics (as in MCMC methods) and rules recede in the modeling
infrastructure. This leads to less parameter-hungry modeling showing a
structured interface to statistical mechanical and machine-learning
techniques - and therefore, perhaps, to robust modeling with stronger
explanatory power.
Vincent Danos, University of Edinburgh