The identification of transcriptional regulatory networks from the vast amount of gene expression data and protein-DNA affinity data is still an open problem. Lately, much effort has been dedicated to the definition of computational and mathematical methods that would be able to shed light into global regulatory patterns that ultimately characterize many phenotypical traits of living beings. These attempts have been successful in providing a wide range of approaches to tackle this identification problem (Herrgård et al., 2004; Ideker et al., 2002) but have so far individually obtained limited success in providing a standard and generally accepted framework for modeling and numerically evaluating gene regulatory networks.

A number of authors have developed models based on the discretization of assumed stated of genes and transcription factors (Akutsu et al., 2000), while others resorted to time dependent material balances (Gupta et al., 2005; Thomas et al., 2004) on mRNA, proteins and other relevant chemical species for the regulatory processes in question. These approaches are inherently valuable for being able to promote advancements on the interpretation of the modeled processes. However, simplifying assumptions in which they are based and/or burdensome experimental requirements prevented them from bring adopted on a general scale. Arguably, another important issue that prevented the dissemination of specific approaches to regulatory patterns identification is their non-straightforward implementation. It is our understanding that tackling transcriptional regulatory network identification problems with the implementation of sound and well-founded applied mathematics frameworks, rather than individualized algorithms designed for specific problems, is crucial to reproducibility and further development of the field.

Under this scope, we have previously formulated an Integer Linear Programming approach based on logical constraints that attempted to reconstruct transcriptional regulatory networks from expression data and location analysis experimental data. We present now improvements on our previous formulation and treatment of the logical relationships that correlate groups of genes and transcription factors. In our proposed framework, these Boolean functions associate each variable state to specific logical relationships between existing states. These logical functions were expanded to take into consideration simultaneous transcription factors regulating the expression of a gene. Finally, the methodology is implemented and evaluated in small and large-scale regulatory networks in Saccharomyces cerevisiae. Results are analyzed for physiological consistency and are further contrasted to previous results obtained for these systems with the use of simplistic logical formulations.

_____________________

Akutsu, T., Miyano, S. and Kuhara, S.; Inferring qualitative relations in genetic networks and metabolic pathways. Bioinf., vol. 16, no. 8, pp. 727-734, 2000.

Gupta, A., Varner, J.D. and Maranas, C.D.; Large-scale inference of the transcriptional regulation of Bacillus subtilis. Comp. & Chem. Eng., vol. 29, pp. 565-576, 2005

Herrgård, M.J., Covert, M.W. and Palsson, B.Ø.; Reconstruction of microbial transcriptional regulatory networks. Curr. Op. in Biotech., vol. 15, pp. 70-77, 2004.

Ideker, T., Ozier, O., Schwikowski, B. and Siegel, A.F.; Discovering regulatory signaling circuits in molecular interaction networks. Bioinf., vol. 18, Suppl. 1, pp. S233-S240, 2002.

Thomas, R., Mehrotra, S., Papoutsakis, E.T. and Hatzimanikatis, V.; A model-based optimization framework for the inference on gene regulatory networks from DNA array data. Bioinf., vol. 20, no. 17, pp. 3221-3235, 2004