Back to Colloquia

Physics Colloquium - Friday, April 23rd, 2010, 3:00 P.M.

E300 Math/Science Center; Refreshments at 2:30 P.M. in Room E200

John Wikswo
Vanderbilt Institute for Integrative Biosystems Research and Education
Vanderbilt University

Avoiding the problem of seven: Can computers design and conduct experiments for automated inference of models of cellular metabolic and signaling networks?

The complexity of biological systems arises from highly nonlinear structural, metabolic, and signaling networks that span multiple scales in space and time. Modern biology is close to providing the complete, reductionist parts list for many simple biological systems, enabling us to address the much harder problem of discerning how all these pieces interact complexity -- a mathematical model of a functioning animal might require Avogadro's number of partial differential equations, termed a Leibnitz. The "problem of seven" may limit the ability of a single human brain to design, conduct and interpret experiments on such complex, non-linear systems. We are developing an integrated measurement and modeling system in which a computer specifies an experiment on isolated cells, the dynamic responses of the cells to a controlled stimulus are recorded using multiple real-time analytical techniques, and the computer then uses these data to select among possible models of the system and propose the next experiment for further model refinement. We demonstrate the ability of Lipson and Schmidt's non-linear regression/estimation-exploration algorithm to infer without a priori assumptions a seven-variable model of metabolic oscillations in yeast. We have yet to determine how far this approach can extend the human seven-variable limit.