The latter is referred to as an “adaptive response”, since adaptation is attributed to reduced damages as a result of adding the priming dose. Compound Library Consequently, it is our goal to characterize and differentiate induced perturbations in terms of the shape and number of computed templates, architecture of the wiring diagrams, and biological interpretation through enrichment analysis. From the perspective of a strict gene expression, the fold changes are generally low and appear to be stochastic as a result of ionizing radiation. This observation is consistent with previous literature. Nevertheless, the temporal patterns from the gene expression provide more candidates and are more informative than a single time point observation, i.e., any transcript with a small value, at a given time point, can be eliminated using standard filtering techniques. The richness of the temporal gene expressions is crucial in grouping and hypothesizing causal relationships from high dimensional transcriptome data. Typically, inference of the causal relationships can be ambiguous; there is significant literature in support of it and against it, but most researchers agree that through carefully designed experimental data, ambiguities in the inference of causation can be reduced or eliminated. Such an experimental design may include a specific set of perturbations that may also include the time-course data. The time-course enables identification of a set of similar profiles that will reduce complexities in the causal network, provide pseudo replicates for sampling and cross validation, and constrain the network structure by enforcing temporal continuity. In short, the proposed computational protocol enables interpretation of a complex dataset at multiple steps. However, the main theme is inference of the simplest network that is computationally tractable, and at the same time, interpretable. The method is initiated by identifying temporally co-regulated transcripts into a distinct set of templates or groups. This step not only reduces the dimensionality of the data, but also reduces the number of variables that need to be estimated for building the causal network, i.e., transition matrices. The network construction assumes a model for which every node, at a given time point, is a sparse linear combination of nodes in the previous time point. The concept of sparseness also enforces the notion of network simplicity. Finally, the solution is regularized by eliciting continuity of the transition matrices between consecutive time points. It should be noted the method has been applied to transcriptome data, but it is also extensible to other time-course data, i.e., identifying aberrant signal transduction pathways. The method has been validated on synthetic data and then applied to transcriptome data that has been collected from a cell strain, which was exposed to 2 Gy ionizing radiation with and without the priming dose of 10 cGy applied 4 hours prior to the higher dose of radiation. Bioinformatics analyses revealed that computed templates without the priming dose are a subset of those that received the priming dose. Furthermore, the adaptive response group included templates with delayed activations and oscillatory behavior. It is clear that the priming dose has induced a significant amount of diversity in how the networks are modulated. In both treatment groups, the initial active templates of the causal networks are highly enriched by the down-regulation of the cell cycle machinery. However, in the case of the adaptive causal network, the network is also modulated by the up-regulation of the inflammatory processes.