Abstract
A comparative interaction detection paradigm is proposed to study the complex gene regulatory networks that control cell proliferation during development. Instead of attempting to reconstruct the entire cell cycle regulatory network from temporal transcript data, differential interactions - represented by generalized logic - are detected directly from time course transcript data under two distinct conditions. This comparative approach is scale- and shift-invariant and is capable of detecting nonlinear differential interactions. Simulation studies on E. coli circuits demonstrated that the proposed comparative method has substantially increased statistical power over the intuitive reconstruct-then-compare approach. This method was therefore applied to a microarray experiment, profiling gene expression in the fruit fly wing as cells exit the cell cycle, and under a condition which delays this exit, over-expression of the cell cycle regulator E2F. One statistically significant differential interaction was identified between two gene clusters that is strongly influenced by E2F activity, and suggests the involvement of the Hippo signaling pathway in response to E2F, a finding that may provide additional insights on cell cycle control mechanisms. Furthermore, the comparative modeling can be applied to both static and dynamic gene expression data, and is extendible to deal with more than two conditions, useful in many biological studies.
