eighty-twenty

I'm sitting in the foyer of the Royal Society in London. This is the venue for Concur 2004, a conference on Concurrency Theory. I'm just attending a couple of the workshops — specifically BioConcur '04 and, next Saturday, AVoCS '04.

BioConcur

So far there's been a talk by Corrado Priami giving an overview of the application of concurrent languages (such as the π-calculus) to the field of Systems Biology.

One success-story so far is that of a study that was made of inflammatory processes in Multiple Sclerosis-affected brain blood vessels. A π-based model was constructed, and simulations of the model were observed. The graphs of the simulations led to the formation of a hypothesis, previously unstudied, relating the number of rolling lymphocytes to blood vessel diameter. Later investigation in mice demonstrated the validity of the hypothesis.

I made a few random notes:

• How does π improve over CCS with regard to cost? Why are names interesting from a cost point-of-view in biological models?
• ODEs (Ordinary Differential Equations) are noncompositional; π is linguistically compositional; it seems to me that these biological models are partly compositional, partly noncompositional, because of the need for encoding of biochemical behaviours (?).
• The table Priami presented relating π to biological systems:

  Molecule	Process
  Interaction Capability	Channel
  Interaction	Communication
  Modification	State and/or channel change

  Compartments are modelled with restriction, in the absence of ambients or ambient-like structures. This is an acknowledged problem with using π for these models.
• The π-based model treats a molecule as a black box with a number of interaction capabilities, each of which has a defined syntactic and behavioural specification for its interactions.
• The exponential distribution used for selecting the next action gives a memoryless property to the system
• In the stochastic π calculus, there is a forced zero collision property — the system excludes the possibility of a collision in competing for resources. This keeps the world nice and deterministic...
• There are standard tools for performance analysis of Continuous Time Markov Chains (link).
• RCCS is another way of modelling these systems - the R stands for "reversible"!
• Model checking is complicated by the stochasticness of the system; model checkers usually like a deterministic universe...
• What kinds of equivalence between processes are biologically interesting?
• Inductive Logic Programming is another alternative approach.