CIDER: a pipeline for detecting waves of coordinated transcriptional regulation in gene expression time-course data

Abstract

Cell adaptability to environmental changes is conferred by complex transcriptional regulatory networks, which respond to external stimuli by modulating the expression dynamics of each gene. Hence, deciphering the network of transcriptional regulation is remarkably important, but proves to be extremely challenging, mainly due to the unfavorable ratio between the number of available observations and the number of parameters to estimate. Most of the existing computational methods for the inference of transcriptional networks consider steady-state gene expression datasets, and produce models of transcriptional regulation best explaining the observed static gene expression.\n\nGene expression time-courses are an emergent typology of gene expression data, paving the way to the characterization of the time-dependent dynamics of transcriptional regulation.\n\nIn this work we introduce the Complexity Invariant Dynamic Time Warping motif EnRichment (CIDER) analysis, a novel computational pipeline to identify the prominent waves of coordinated gene transcription induced in cells by external stimuli, and determine which TFs are involved in the coordination of gene transcription. The CIDER pipeline combines unsupervised time series clustering and motif enrichment analysis to first detect transcriptional expression patterns, and then identify the TFs over-represented in the promoter regions of gene sets with similar expression dynamics.\n\nThe ability of CIDER to correctly identify regulatory interactions is assessed on a realistic synthetic dataset of gene expression time-courses, generated by simulating the effects of knock-out perturbations on the E. coli regulatory network.\n\nThe CIDER source code and the validation datasets are available on request from the corresponding author.