UC San Diego

Organisms use a complex transcriptional regulatory network (TRN) to sense their environments and alter their phenotypes in response, which is integral to their adaptation and survival. Transcriptomic datasets, which measure the complete activity of the TRN as expression values for each gene, have accumulated online in large numbers, and represent a goldmine of biological information. However, the large number of variables and uncharacterized global TRN structure present a significant challenge for interpreting these datasets. A recently developed machine learning method, iModulon analysis, can determine the global structure of TRNs by identifying co-regulated signals in compendia of diverse transcriptomic data. The output of this method is a set of independently modulated gene sets (iModulons) for the given organism, which can be characterized to discover transcriptional regulation mechanisms and other biological insights. Characterized iModulons enable a low dimensional, knowledge-enriched representation of the transcriptome, dramatically simplifying the analysis of the sensory and regulatory systems of cells. Here, we establish and expand the usefulness of iModulons in three aims: (1) we characterize the TRN of Bacillus subtilis, a model soil and gut bacterium, obtaining novel insights on a variety of functions including sporulation; (2) we develop iModulonDB.org, and online knowledgebase of iModulons, and we populate it with over ten characterized organisms from across the phylogenetic tree of life to create a widely applicable scientific knowledgebase; and (3) we combine iModulon analysis with DNA sequencing and other technologies to understand the evolution of stress tolerance for both oxidative and temperature stress. This body of work stands upon decades of bottom-up characterization of transcriptional regulators and analyzes years’ worth of accumulating datasets to elucidate a new perspective on cellular function from the top-down. It reveals that a global understanding of the transcriptome in low dimensions is both possible and useful, and provides an online resource so that the scientific community can continue to mine it for new insights. Stress resistance mechanisms presented in Aim 3 are relevant to the evolution of pathogens and common biomanufacturing challenges. Taken together, this work establishes iModulons as a powerful and accessible tool for the understanding of gene expression.