Figure 1.
Relationship between mean and variance in protein expression.
a) Protein mean and variance values in S. cerevisiae plotted against each other in log-scale in arbitrary fluorescence, with corresponding Pearson's correlation coefficient. b) Distribution of residual variance values across the S. cerevisiae dataset. Red bars indicate residual variance value with Z-scores over 2 standard deviations from the mean.
Figure 2.
Stochastic model of gene expression: a) Schematic representation of the model.
Each step transition is determined by a rate constant. Promoter activation and inactivation occur at Kon and Koff rates respectively. When active, a promoter is transcribed at Km rate into an mRNA molecule. The mRNA molecule can then be either degraded at Dm rate or translated at Kp rate into a protein. The protein molecule can then be degraded at rate Dp. Kon, Koff, and Km determine the synthesis rate of mRNA, or Sm. Blue indicates that the parameter has been empirically measured or calculated across the dataset, red indicates that the parameter has been simplified or fit across the dataset b) Model performance in predicting protein variance in S. cerevisiae. Each point represents a single GFP fusion strain. Data is displayed in log-scale (linear scale r = 0.836).
Figure 3.
Promoter kinetics but not mRNA and protein synthesis and degradation rates modulate the relationship between mean and variance.
a) Predicted relationship between mean and variance using original model with original parameter set (grey squares), original model with permuted sets of kinetic rates for mRNA/protein synthesis and degradation (purple), and slow promoter kinetics model with original parameter set (orange). b) Fraction of residual variance explained (r2) by sources of noise operating at the promoter/initiation level (orange) or at a post-initiation level (purple).
Figure 4.
Analysis of mRNA distributions connects underlying promoter kinetics to nucleosome occupancy.
a) mRNA mean and variance in S. cerevisiae plotted against each other in log-scale. Blue dashed line indicates the expected relationship between mean and variance in a regime of slow activation and fast inactivation rate (σ2 = μ), red dashed line indicates expected relationship at slow promoter kinetics (σ2 = μ+μ2). Circles represent experimental values of mRNA mean and variance (color matches best fit to promoter kinetics regime) b) Average nucleosome occupancy between −600 to +1000 relative to the TSS of S. cerevisiae genes exhibiting linear mRNA mean-variance scaling. The position of the canonical nucleosome free region is indicated by the black arrow. c) Same as b) but with respect to S. cerevisiae genes exhibiting quadratic mRNA mean-variance scaling.
Figure 5.
Modulating promoter kinetics changes protein mean-variance scaling.
a) Increment in residual variance from glucose to ethanol in genes that show increased occupancy in ethanol (orange set: test) and genes with unaltered occupancy (orange set: control) compared to the same genes ranked by high (purple set: test) or low (purple set: test) increase in translation rate (purple set) (* indicates p<0.05, t-test). b) Diagram connecting the power-law exponent to promoter kinetics: most genes in S. cerevisiae exhibit promoter kinetics characterized by fast inactivation rate (purple dots) and display protein mean-variance scaling dictated by a power-law with 1.69 exponent (purple line). A small set of genes (orange dots) exhibit slow promoter kinetics and consequently present protein mean-variance scaling dictated by a quadratic scaling (orange line).