Graphical models for reasoning on biological systems: computational challenges
Satellite meeting of ECCS'10 (European Conference on Complex Systems).
September 16, 2010, Lisbon, Portugal
9:00 - 10:00 1st session (1h) :
9:00 - 9:15 : introducing the workshop
9:15 - 10:00 : Invited speaker Barnes and Stumpf (Imperial College, London, UK) "Approximate Bayesian Computation (ABC) to Learn the Structure and Dynamics of Complex Systems"
10:00 - 10:30 coffee-break
10:30 - 12:30 2nd session (2h)
11:30 - 12:00 : Gay et al. "A Graphical Method for Reducing and Relating Models in Systems Biology"
12:30 - 14:00 lunch
14:00 - 16:00 3rd session (2h)
14:00 - 14:45 : Invited Speaker Janine Illian (University of Saint-Andrews, UK) "Ecological applications of spatial point process theory - examples of spatial complexity"
15:25 - 15:55 : Baudrit et al. " Hybrid parameter learning using Dirichlet distributions for modelling the dynamics of multi-scale phenomena occurring in food processes"
15:55 - 16:00 : Wrap up
A modelling drill for scientists is now to build up models that process modern biological data in order to gain a better understanding of involved mechanisms or for control/management of biological systems. These data are characterized by a large number of entities in interactions (e.g. individuals in ecology, hosts in epidemiology, genes in regulation networks). Graphical models incorporate an interesting theoretical framework in terms of flexibility to integrate such interactions and an intuitive representation of the system. Still new computational issues arise when designing graphical models specific to the data we just mentioned.
In genomics and bioinformatics, biological network modelling aims at describing interactions between the components present in a cell. These interactions may furthermore evolve along time. Undergoing processes or structures can often be described via latent variables. Models combining all these dimensions raise strong difficulties in terms of inference, mainly because of very intricate dependency structures. Some distributions can then only be approximated and classical statistical methodology therefore needs to be adapted.
In epidemiology or ecology, the systems under study present not only spatial but also temporal dynamics. The number of nodes (individuals) can be moderate as compared to systems biology problems, but dealing with the interactions is still challenging. If the biological system is controlled or managed (epidemics control, species conservation), reasoning on these highly structured systems is limited by the capacities of the available decision optimization frameworks. Due to all these specificities, it is necessary to adapt classical methodological tools for estimation/prediction/optimization/planning.
The aim of this meeting is to discuss how reasoning on or controlling biological systems using graphical models raise new computational issues and what are the current research trends to solve these issues.
We are interested in applications from the bioinformatics, epidemiology and ecology domains. We do not limit the kind of biological systems studied: interactions can be the results of spatial and/or temporal dynamics, of social behavior, of causality … Both graphical models of stochastic or deterministic (e.g. Constraint Satisfaction Problems and thier extensions) nature can be fully relevant. Computational issues discussed can be relevant to parameters estimation, structure inference, optimization for decision under uncertainty, etc.
The expected outcome of the workshop will be to understand how biological applications/data do renew the classical framework of graphical models and to highlight some breakthroughs that were explicitly developed in this area.
In the proposed meeting, we would like to leave the door open to scientists from both (i) computational science - computer scientists, mathematicians (including modellers, statisticians, machine learning or artificial intelligence persons, operations research people, etc.) interested in tackling complex biological issues by developing graphical/computational solutions to represent the system and the knowledge contained in the data and (ii) biology - scientists who feel the need for such computational approaches and would like to know whether a complex systems view could help them answer newly raised biological issues.
Call for paper
Two types of submissions are welcome. Short papers concerning work in progress or presentation of a methodological issue are encouraged. This can typically be a starting PhD work, but short papers are not restrained to that. Longer papers should present more mature works and results.
Short papers length is limited to 6 pages maximum, long papers should be no more than 12 pages long. The document format template proposed by ECCS for submission to the conference should be used for all submissions. They are available here. Contributions can be submitted no later than 30th April 2010 sending us an email (see contacts below).
- 10th of Feb. 2010, call for papers opened.
- 30th of April 2010, deadline for the submission of contributions to the satellite meeting. Now extended to 25th of May 2010
- 31th of May 2010, notification of paper acceptance to the meeting. Now extended to 21th of June.
- 12th of June 2010, deadline for submission of final versions of contributions. Now extended to 7 th of July.
- 13th to 17th of Sept. 2010, ECCS'10 conference.
- 16th of Sept. 2010, ECCS Satellite Meeting "Graphical models for reasoning on biological systems: computational challenges".
(Satellite Meeting attendees must register for the main ECCS’10 conference.)
- Nathalie Peyrard (BIA, INRA Toulouse, France)
- Stéphane Robin (AgroParisTech, Paris, France)
- Régis Sabbadin (BIA, INRA Toulouse, France)
- Matthieu Vignes(BIA, INRA Toulouse, France)
For paper submission or enquiries about the meeting, please contact the organisation committee at email@example.com