Data-driven identification of inherent features of eukaryotic stress-responsive genes

Virtually all living organisms are challenged by changes in their environment, such as sudden increases in osmolarity. To survive, cells have developed stress-responsive mechanisms that tune functions such as metabolism and gene expression. The response of Saccharomyces cerevisiae to osmostress includes a massive reprogramming of gene expression. Identifying the inherent features of stress-responsive genes is of significant interest for understanding the basic principles underlying the rewiring of gene expression upon stress. Here we generated a comprehensive catalog of osmotress-responsive genes from 5 independent RNA-seq experiments. We explored 30 features of yeast genes and found that 25 (83.34%) were distinct in osmostress-responsive genes. We then identified a subset of 13 non-redundant minimal osmostress gene traits and used statistical modeling to generate a ranking of the most stress-predictive features. Our findings reveal that the most relevant features of osmostress-responsive genes are the number of transcription factors targeting a gene and gene conservation. The same features that define yeast osmostress-responsive genes can also be used to predict osmostress-responsive genes in humans, although the most predictive features in humans are GC content and mRNA half-life. Our study provides a holistic understanding of the basic principles of the regulation of stress-responsive gene expression across eukaryotes. Overall design: RNA-seq gene counts from yeast cells (HOG1-WT or HOG1-KO) challenged with hypertonic stress (0.4 M NaCl, 15 min) or kept in regular medium. Each condition in triplicates. 18 samples.