Systems Biology

Network Failure

+ See all authors and affiliations

Science  07 Nov 2008:
Vol. 322, Issue 5903, pp. 823-825
DOI: 10.1126/science.322.5903.823d

Models of metabolic and signalling networks have been characterized, perhaps unfairly, as reannotations of previously discovered interactions. To counter this concern (and the statistical issue of sorting through hundreds of correlations), Janes et al. describe an approach called “model breakpoint analysis” that stresses the network by using nonphysiological inputs in a manner similar to that of engineers performing failure analysis of bridges or cars. They began with their model of cytokine-induced apoptosis and proceeded to introduce implausible data that stretched the dynamic range of the cell (defined as the responsiveness of cell outcomes to incremental changes in cell activation). Surprisingly, network function did not degrade in parallel, but worked perfectly well until a threshold (or breakpoint) was reached, at which point the predictions were no longer useful. Pinpointing the signals and stimuli that were responsible for the system failure enabled them to distinguish epiphenomena from causal factors and to make predictions about the dynamic roles of three kinases (Akt, ERK, and Mk2) in cytokine-induced apoptosis. These predictions were then confirmed in inhibitor- and mutant-based experiments, suggesting that differences in dynamic range can be more important to cellular function than the strength of a particular signal. — BJ

Cell 135, 343 (2008).

Navigate This Article