Transcriptional interactions are one of the important components which govern gene expression. Their ability to alter gene expressions is dependent primarily upon the amount of activated transcription factor present in the system as well as the binding kinetics of said transcription factor. Therefore, the overall mechanism of transcriptional activation is that some ligand will activate a given transcription factor, this transcription factor will translocate into the nucleus, and then either up or down-regulate a given gene. The complexity of the system arises from the fact that transcription factors as well as their ligands can themselves be protein. Therefore, the activation of a given transcription factor may in turn alter the production of other transcription factors, which may eventually feedback and regulate the initial transcription factor via various means. Therefore, complexities arise from such a system because the activated levels of a given transcription factor may be regulated by the levels of many other transcription factors.

Thus, rather than thinking of transcriptional interactions as merely the interactions of an activated transcription factor with their various gene targets, it makes more sense to think of the interactions as a complex network in which transcription factors can interact with each other. However, to obtain such a network is important to quantify how the activation of one transcription factor can affect the level of activation of the other transcription factors.

The Living Cell Array (LCA)[1] is a novel microfluidics device which is able to determine the levels of an activated transcription factor in a high throughput manner. Not only is it able to obtain the levels of a set of transcription factors in parallel, but it is able to do this with a high temporal resolution as well under multiple stimulatory conditions. It accomplishes this by first utilizing a highly specific marker for the activity of a specific transcription factor. It accomplishes this task by first utilizing a plasmid which contains an unstable green fluorescence protein (GFP), a minimal promoter, and 4 repeats of a transcription factor consensus sequence obtained via TRANSFAC[2] which is highly specific for a given transcription factor.

When a transcription factor is activated by its specific external factor, it will then bind to the binding regions in the plasmid which then causes the transcription of the unstable GFP. This GFP due to its unstable nature is able to reflect the amount of transcription factor present at a relatively instantaneous point in time. Thus the fluorescence detected in a given cell functions as an accurate surrogate for a given transcription factor. The use of the microfluidics device then allows populations of identical cells to be stimulated in parallel with different stimulatory cocktails as well to be imaged in a high throughput manner.

The current experimental implementation of the LCA was used to explore first the response of hepatic cells to various inflammatory cytokines such as TNF-alpha, IL1, IL6, IFG-gamma, and the anti-inflammatory drug Dexamathasone. To quantify the effect of these factors upon inflammation, the transcription factor NFkB, STAT3, STAT1, HSE, GRE, ISRE were analyzed.

Because of the structure and quality of the data obtained via the LCA, it open up new opportunities for computational analysis, specifically through the determination of both network architectures as well as various clues as to the underlying mechanisms associated with transcription factor activation. We propose the use of a novel numerical technique to analyze the LCA data which we term Reverse Euler Deconvolution (RED)

Like many techniques, RED utilizes the general hypothesis that the system can be modeled via a set of ordinary differential equations in which the effect of multiple transcription factors can be treated in an additive sense. However, because of the nature of the data, in which there are more stimulatory profiles than there are transcription factors as well as the high temporal resolution of the data, we do not have to impose constraints to make the problem well-posed. In contrast to other techniques such as NCA[3], NIR[4], PLS[5], RED coupled with the LCA does not need to enforce any type of network architecture upon the data in the form of specific conditions a network must meet or in the number of connections that is required to fully define the system. Furthermore, RED treats the matrix A not as a matrix of constants as in the previous cases, but as time varying functions whose numerical response may provide insights as to the underlying process.

Utilizing a mixed integer linear programming formulation, we are not only able to solve for the time varying responses which are most likely with a given network architecture, but also the network architecture itself. Furthermore, we can show that this network architecture illustrates a significant conservation of interactions over multiple solutions as well as the fact that the interactions themselves show characteristic behavior of simple network motifs in response to a step input.

1. King KR, Wang S, Irimia D, Jayaraman A, Toner M, Yarmush ML: A high-throughput microfluidic real-time gene expression living cell array. Lab Chip 2007, 7:77-85.

2. Matys V, Fricke E, Geffers R, Gossling E, Haubrock M, Hehl R, Hornischer K, Karas D, Kel AE, Kel-Margoulis OV, et al: TRANSFAC: transcriptional regulation, from patterns to profiles. Nucleic Acids Res 2003, 31:374-378.

3. Liao JC, Boscolo R, Yang YL, Tran LM, Sabatti C, Roychowdhury VP: Network component analysis: reconstruction of regulatory signals in biological systems. Proc Natl Acad Sci U S A 2003, 100:15522-15527.

4. Gardner TS, di Bernardo D, Lorenz D, Collins JJ: Inferring genetic networks and identifying compound mode of action via expression profiling. Science 2003, 301:102-105.

5. Boulesteix AL, Strimmer K: Predicting transcription factor activities from combined analysis of microarray and ChIP data: a partial least squares approach. Theor Biol Med Model 2005, 2:23.