Patterns of extreme outlier gene expression suggest an edge of chaos effect in transcriptomic networks

Most RNA-seq datasets harbor genes with extreme expression levels in some samples. Such extreme outliers are usually treated as technical errors and are removed from the data before further statistical analysis. Here we focus on the patterns of such outlier gene expression to investigate whether they provide insights into the underlying biology.

Our study is based on multiple datasets, including data from outbred and inbred mice, GTEx data from humans, data from different Drosophila species, and single-nuclei sequencing data from human brain tissues. All show comparable general patterns of outlier gene expression, indicating this as a generalizable biological effect. Different individuals can harbor very different numbers of outlier genes, with some individuals showing extreme numbers in only one out of several organs. Outlier gene expression occurs as part of co-regulatory modules, some of which correspond to known pathways. In a three-generation family analysis in mice, we find that most extreme over-expression is not inherited, but appears to be sporadically generated. Genes encoding prolactin and growth hormone are also among the co-regulated genes with extreme outlier expression, both in mice and humans, for which we include also a longitudinal expression analysis for protein data.

We show that outlier patterns of gene expression are a biological reality occurring universally across tissues and species. Most of the outlier expression is spontaneous and not inherited. We suggest that the outlier patterns reflect edge of chaos effects that are expected for systems of non-linear interactions and feedback loops, such as gene regulatory networks.