Skip to main content
Fig. 3 | Genome Biology

Fig. 3

From: Selected heterozygosity at cis-regulatory sequences increases the expression homogeneity of a cell population in humans

Fig. 3

Correlation of selection for heterozygotes with cell-to-cell variation in gene expression and chromatin accessibility. a Illustration of how heterozygosity can buffer stochastic noise caused by the fluctuations of binding regulators. Red and blue spheres indicate TFs binding to the G and A alleles, respectively. Variation in the blue RNAs directly reflects variation in the blue TFs across the four homozygous cells (top), whereas variation in the red RNAs compensates for variation in the blue RNAs among the heterozygotes (bottom). It is assumed that the sources of extrinsic noise affecting the two alleles are uncoupled. b Our mathematical model detailed in the “Methods”. K 1 is the dissociation constant between the promoter and RNA polymerase (R). K 2 is the dissociation constant between the cis-regulatory region and TF. For repressing TFs, K 3 is used in place of K 1. The noise levels in protein expression are compared between the heterozygote with (K 2, λK 2) and the homozygote with \( \left(\overline{\lambda}{K}_2,\overline{\lambda}{K}_2\right) \), given the same level of average gene expression. c The differences in the squared coefficient of variation (CV2) between the heterozygote and homozygote were obtained as a function of the allele discrepancy parameter λ for the varying TF concentration parameter a. As λ deviates from one, the two alleles of the heterozygote are more differentially regulated. Whether the TF is an activator (top) or repressor (bottom), the noise differences ∆η 2 were constantly greater than zero, indicating that the homozygote introduces higher levels of transcriptional noise than the heterozygote. d Gene expression level as a function of the Tajima’s D of the associated regulatory SNP. RNA sequencing data in four white blood cells from the Roadmap Epigenomics project (top) and DNA microarray data in GM12878 before and after particular treatments (bottom) were normalized (see “Methods”) before plotting. e Progressive reduction of the expression noise (CV2) in proportion to the Tajima’s D of the associated regulatory SNP. The curves show a fit to RNA sequencing data for 62 single cells with the solid and dashed lines representing heterozygous and homozygous loci in the GM12878 cells, respectively. The heterozygous curves are divided based on Tajima’s D. The green dots indicate representative genes that have a detectable RNA sequencing measure and are mapped to footprint SNPs heterozygous in GM12878 with D > 1.5. f Observed cell-to-cell variability in the chromatin accessibility of the cis-regulatory regions categorized in the same manner as in e. Error bars represent one standard deviation of the variability obtained through bootstrapping (see “Methods”)

Back to article page