A mutual information approach to reverse- engineering plastic transcription regulatory networks in Arabidopsis thaliana under differ- ent abiotic stresses

ARCE, AGUSTÍN L.; CHAN, RAQUEL L.; CHERNOMORETZ, ARIEL

Ciudad Autónoma de Buenos Aires

Workshop; XV Giambiagi Winter School; 2013

Departamento de Física, Facultad de Ciencias Exactas y Naturales, UBA

Algorithms developed for the reverse engineering of gene regulatory networks (GRNs) generally require large datasets which comprise various cell types and treatments. As a trade-off, they obtain "aggregative" static GRNs. However, the activity of most genetic interactions has a great spatiotemporal dependency and can be altered by external stimuli. Plants constitute a very attractive system for reverse-engineering GRNs. Their sessile nature has been evolutionary compensated by a remarkable phenotypic plasticity in response to external stimuli. Accordingly, biotic and abiotic stresses cause large transcriptomic alterations. In order to study abiotic stress GRNs, the AtGenExpress abiotic stress dataset was selected. Samples include a control and nine abiotic stress treatments. According to sample source, microarrays were divided in two subsets, roots and shoots. They were filtered assessing microarray quality and informative gene-expression profiles. Then, the ARACNe algorithm was employed to obtain the transcription regulatory networks (TRN). A plot of the PCC between linked genes versus their mutual information (MI) (Fig. 1a) shows how ARACNe recovers putative nonlinear interactions. The activity state of the links under the different treatments was determined evaluating the PCC for each subset of measures (Figure 2b). This allowed the recognition of treatment-specific TRNs and their functional and comparative analysis. For example, a multidimensional scaling study showed that some related stresses, as osmotic, salt and cold treatments, present similar TRNs in roots. In addition, interactions active across all treatments where identified recovering consitutively active TRNs (Figure 1c). The functional analysis of the genes included indicated a cross-talk between circadian and abiotic stress genes in shoots, which could mediate the known regulation of these responses throughout the day. The approach described allowed the reconstruction of plastic abiotic TRNs enabling the characterization and comparison of treatment-specific and constitutive responses.