Thousands of Arabidopsis thaliana transcript profiling studies report transcript dynamics based on abiotic and biotic stresses, chemical treatments, or development. Organizing these datasets could reveal the structure of responses and cross-talk, and in which cells the plants perceive, signal, respond to and integrate environmental inputs.

We have clustered Arabidopsis transcript profiles for >22,000 genes and >150 treatments, comprising abiotic, biotic and chemical stresses. A new clustering procedure resulted in ~180 clusters that explain the response dynamics of the Arabidopsis genome. In particular, we identify ubiquitous stress responses in Arabidopsis - similar to those of fungi and animals - that employ genes in pathways related to Snf1- or MAP-kinases, vesicle transport, mitochondrial functions, and the transcription machinery. The ABA-dependent transcriptome is clearly delineated in well-defined clusters, while functions dependent on reactive oxygen species are widely distributed, possibly indicating evolutionary pressures conferring distinction to different stresses in time and space.

In further studies, the Graphical Gaussian Model (GGM) was used to assemble a gene network for the Arabidopsis transcriptome. Based on partial correlation (pcor), GGM infers co-regulation patterns between gene pairs conditional on the behavior of other genes. We used "regularized" GGM, coupled with iterative random samplings, to expand the network to cover the entire Arabidopsis genome. Several network variants include up to 50% of the Arabidopsis transcriptome. When querying for selected genes, locally coherent sub-networks emerge that are often related to metabolic functions and stress responses. GGM recovers interactions with biological significance that typically escape capture by Pearson correlation networks. Finally, the network reconciles individual sub-networks in a topology joined at the whole genome level, and provides a general framework that can instruct future studies on plant metabolism and stress responses.

The Graphical Gaussian Model software is available from from Genome Research (supplemental data).