Modeling of Stress-induced Regulatory Cascades Involving Transcription Factor Dimers

Regulatory cascades consisting of stress-induced gene modules and their transcriptional regulators were recently identified and quantitatively modeled using Artificial Neural Networks (ANNs). Here, we extent our approach to account for regulators that consist of transcription factor dimers. The latter are frequently missed by module-finding tools since the expression profile of one of the dimers is typically un-altered while the profile of the second dimer changes significantly during the stress response. We identify two modules of stress-associated genes that are regulated by transcription factor dimers and show that ANN modeling can be used to accurately predict the expression of all genes in these modules. We also find that prediction accuracy is dependent on the specific stress category, in agreement with experimental studies where constitutively expressed transcription factors exert different regulatory actions upon different exposures to stressful stimuli. Overall, we show that TF dimerization is an important attribute of the transcriptional network regulation, which is frequently ignored by module finding tools, and can lead to the identification of important biological cascades whose regulatory action can be quantitatively simulated with ANNs.