Relationships between gene expression variability, expression levels, and Protein–Protein interactions in mouse and yeast

Understanding how evolution shapes variability in gene expression is complicated by the interdependence of expression levels, expression variability, and evolutionary rates. Metrics that quantify variability independently of expression would help disentangle these relationships. Previously, a metric termed F* was developed using single-cell RNA-seq data from Mus musculus. Here, analyses of single-cell RNA-seq data from M. musculus and Saccharomyces cerevisiae reveal that the relationship between expression levels and variability is more complex than expected, and F* cannot be fully separated from expression level. Comparisons between single-cell and non-single-cell or simulated bulk experiments show that single-cell data exhibit higher apparent variability for most genes, consistent with contributions from both intrinsic and extrinsic sources. Despite this, the negative relationship between protein-protein interaction connectivity and F* is conserved in both organisms and is also detectable in some non-single-cell datasets, indicating it is not unique to single-cell data or variability. Gene ontology analyses show that, in M. musculus across single-cell and non-single-cell datasets, low-F* genes are enriched for translation- and ribosome-associated functions. When single-cell-specific variability is isolated by controlling for non-single-cell contributions, additional enrichment emerges for splicing and spliceosome-associated genes in M. musculus, suggesting that genes encoding spliceosome components in M. musculus may be under selective pressure to maintain unusually low variation relative to both their expression levels and the variability observed in bulk systems. In no analyses are there over-represented GO terms among the S. cerevisiae genes with low F* values.