Date: November 26, 2012; 14:00 – 17:30
Venue: Seminar Room #D312, The Institute of Statistical Mathematics, Research and Organization of Information and Systems, JAPAN
Organizer: Research and Development Center for Data Assimilation, The Institute of Statistical Mathematics, JAPAN
Program
14:00 – 14:10 Opening
Tomoyuki Higuchi (The Institute of Statistical Mathematics, Director-General)
14:10 – 15:10 Anomalous thermodynamics at the microscale (invited)
Antonio Celani (Institut Pasteur, FRANCE)
Particle motion at the micro-scale is an incessant tug-of-war between thermal fluctuations and applied forces on one side, and the strong resistance exerted by fluid viscosity on the other. Friction is so strong that completely neglecting inertia – the overdamped approximation – gives an excellent effective description of the actual particle mechanics. In sharp contrast with this result, here we show that the overdamped approximation dramatically fails when thermodynamic quantities such as the entropy production in the environment is considered, in presence of temperature gradients. In the limit of vanishingly small, yet finite inertia, we find that the entropy production is dominated by a contribution that is anomalous, i.e. has no counterpart in the overdamped approximation. This phenomenon, that we call entropic anomaly, is due to a symmetry-breaking that occurs when moving to the small, finite inertia limit. Strong production of anomalous entropy is traced back to intense sweeps down the temperature gradient.
15:10 – 15:20 Break
15:20 – 15:50 Data assimilation on intracellular fluid dynamics in C. elegans embryo
Hiromichi Nagao (The Institute of Statistical Mathematics)
Data assimilation is applied for the first time to the intracellular fluid dynamics focusing on the cytoplasmic streaming in Caenorhabditis elegans embryo, which is observed shortly after the fertilization. The purpose of the present work is to quantitatively obtain the spatial and temporal distributions of the force generated by the molecular motors, which is considered to drive the cytoplasmic streaming. The estimated force distribution that is reconstituted from the maximum-a-posteriori solution, which attains the posterior distribution function maximum, is found no longer to be proportional to the velocity distribution. This fact may indicate strong forces are generated even at region where the streaming is slow.
15:50 – 16:20 Bayesian methods for making systems, molecules and others
Ryo Yoshida (The Institute of Statistical Mathematics; CREST, Japan Science and Technology Agency (JST))
Modeling biochemical kinetic systems based on ordinary differential equations (ODEs) or algebraic equations has been practiced widely in metabolic engineering and many researches on systems biology. This has promoted in-depth understanding of action, design and control of the potentially complex dynamics. The system consists of a network of interacting biochemical species, which involves production, degradation, diffusion and binding of many substances. This study attempts, as the primary focus, to achieve a previously-unexplored task of the system model construction; robust biochemical networks are designed automatically according to the principle of Bayesian robustness. Certain types of biological systems are hypothesized to maintain robustness with respect to specific instances of systems perturbation and environmental uncertainty.
The present Bayesian method is aimed at exploring a robust ODE network such that the model outputs fit given data robustly in response to artificially-created kinetic perturbations, damages, reaction failures and environmental uncertainty. The key idea lies in the use of the Dirichlet process mixture (DPM) as a prior distribution.
Once system perturbations or kinetic parameter variations are modeled by the DPM prior, relevant robust systems can automatically be constructed while in the stochastic graph search process. We provide the general framework for the Bayesian robust networking, technical details on the Markov chain Monte Carlo algorithm, and demonstrations on some benchmark problems.
16:20 – 16:30 Break
16:30 – 17:00 Enhancement of collective immunity by selective vaccination against emerging influenza pandemic
Masaya M. Saito (The Institute of Statistical Mathematics)
Since the stockpile of vaccines that can be prepared before the arrival of an emerging pandemic strain is generally very limited, priorities in vaccinations should be assignedpreferentially to certain specific social groups. We create a priority design for collective immunity and simulate the spread of influenza to determine the target groups for vaccination in a model city. The simulation result shows that the illness attack rate can be reduced from 40% of the baseline cases to 5%, if students and office workers are intensively dosed within three months. This administration design gives lower priority to elderly people, having a higher risk than younger people. Nonetheless, this administration design reduces the mortality rate further across all ages than random administrations to all social groups.
17:00 – 17:30 Statistical approach towards intracellular information processing
Tetsuya J. Kobayashi (Institute of Industrial Science, the University of Tokyo; PRESTO, Japan Science and Technology Agency (JST))
Microscopic biological systems are intrinsically stochastic due to thermal fluctuation and small number of molecules whereas whole biological systems operate robustly. I demonstrate that information theoretical and statistical knowledge can be employed to understand the design principle for emergence of robust operation out of stochastic components in biological systems.
