This article is part of the supplement: Selected articles from the Third International Symposium on Optimization and Systems Biology
Identification of transcription factor's targets using tissue-specific transcriptomic data in Arabidopsis thaliana
1 Computer Science Department and Christopher S. Bond Life Sciences Centre, University of Missouri, Columbia, USA
2 Monsanto Company, Creve Coeur, St. Louis MO, USA
BMC Systems Biology 2010, 4(Suppl 2):S2 doi:10.1186/1752-0509-4-S2-S2Published: 13 September 2010
Transcription factors (TFs) regulate downstream genes in response to environmental stresses in plants. Identification of TF target genes can provide insight on molecular mechanisms of stress response systems, which can lead to practical applications such as engineering crops that thrive in challenging environments. Despite various computational techniques that have been developed for identifying TF targets, it remains a challenge to make best use of available experimental data, especially from time-series transcriptome profiling data, for improving TF target identification.
In this study, we used a novel approach that combined kinetic modelling of gene expression with a statistical meta-analysis to predict targets of 757 TFs using expression data of 14,905 genes in Arabidopsis exposed to different durations and types of abiotic stresses. Using a kinetic model for the time delay between the expression of a TF gene and its potential targets, we shifted a TF's expression profile to make an interacting pair coherent. We found that partitioning the expression data by tissue and developmental stage improved correlation between TFs and their targets. We identified consensus pairs of correlated profiles between a TF and all other genes among partitioned datasets. We applied this approach to predict novel targets of known TFs. Some of these putative targets were validated from the literature, for E2F's targets in particular, while others provide explicit genes as hypotheses for future studies.
Our method provides a general framework for TF target prediction with consideration of the time lag between initiation of a TF and activation of its targets. The framework helps make significant inferences by reducing the effects of independent noises in different experiments and by identifying recurring regulatory relationships under various biological conditions. Our TF target predictions may shed some light on common regulatory networks in abiotic stress responses.