Exploring variation in biochemical pathways with the continuous pi-calculus

Thursday, February 23rd 2012. 4pm, CS1.01

Abstract

Theoretical computer science and biology have common cause in trying to model and understand complex interacting processes. One potential contribution from computer science is the use of high-level languages for concurrent systems to smooth the route between natural descriptions and mathematically precise models. This is similar to the relation between programming languages and compiled executables: these languages allow us to express and communicate intuition, while keeping contact with low-level realities.

The continuous pi-calculus is one such language for modelling biochemical systems. It is based on Milner's pi-calculus, a language of interacting processes with dynamic communication topology. Continuous pi modifies this to express the real-valued concentrations, reaction rates, and multiway interactions of biochemistry. Models in continuous pi are reagent-centred descriptions of individual chemical species, expressing their potential behaviour and interactions. These descriptions can be compiled into ordinary differential equations; and numerical simulation then shows the behaviour over time of biochemical mixtures: reactions, complexes, enzymes, activation and inhibition.

As well as modelling specific biochemical systems, these high-level descriptions allow us to investigate potential variations to those systems, and to explore the accessible evolutionary landscape. To demonstrate this we modelled a classic intracellular signalling pathway and then simulated every variant system reachable by a single change. We then used assertions in temporal logic to evaluate how well these variant pathways carried a signal. This complete survey of the one-step neighbourhood highlights which parts of the pathway are robust, maintaining behaviour under change, and which parts offer potentially new behaviours.

This investigation of system behaviour under variation gives an example of how high-level languages for describing biological processes make it possible to express and test correspondingly high-level hypotheses about robustness, neutrality and evolvability.

Work on continuous pi is a collaboration with Marek Kwiatkowski and Chris Banks.

On Executable Models of Molecular Evolution. Kwiatkowski and Stark. In Proceedings of the 8th International Workshop on Computational Systems Biology WCSB 2011, pages 105–108. (PDF)

Biography

Ian Stark is a Senior Lecturer in Computer Science at the University of Edinburgh School of Informatics, where he works in the Laboratory for Foundations of Computer Science. His research is on mathematical models for programming languages and concurrent interacting systems, in particular reasoning about names, local state, and mobility. His work on biochemical modelling is in association with the SynthSys Centre for Integrative Systems Biology at Edinburgh (formerly CSBE).

Graham Cormode, AT&T Labs Research

Mergeable Summaries

Thursday, March 15th 2012. 4pm, CS1.01

Abstract

In dealing with massive distributed data, exact computations may be possible but can still be very expensive to perform. Instead, it is often desirable to look to more lightweight computations that give (guaranteed) approximations. A central approach is the idea of the mergeable summary: a compact data structure that summarizes a data set, which can be merged with others to summarize the union of the data without significant growth in size. This framework enables parallel computation.

Samples and sketches are two examples of well-known, effective mergeable summaries. I'll present recent results on new, compact mergeable summaries for various tasks, such as approximating frequent items, order statistics, and range counts.

Joint work with Pankaj Agarwal, Zengfeng Huang, Jeff Phillips, Zhewei Wei, Ke Yi

Biography

I am a researcher at AT&T Labs--Research in New Jersey. I work on data stream analysis, massive data sets, and general algorithmic problems.

Before this, I worked at Lucent Bell Laboratories, with focus on Network Management, and previously, I was a Postdoc researcher at the DIMACS research facility, which is located at Rutgers University. I retain connections to DIMACS and to MassDAL -- the Massive Data Analysis Lab. I did my PhD work at the Department of Computer Science at the University of Warwick, UK, and spent some time studying in Cleveland, Ohio at Case Western Reserve University with the Electrical Engineering and Computer Science Department .

Paterson lecture

Iain Stewart, University of Durham

TBA

Thursday, May 10th 2012. 4pm, CS1.01