Figure 3.

Variance Stabilizing Transformation (VST). Panel A depicts the raw difference ${\mathcal{D}}_g=Y_{g1}-Y_{g2}$ for 20,000 simulated genes (Methods Simulations), suggesting that the variance of D_Q increases as the mean µQE increases; hence, there is no uniform cutoff to differentiate DEGs and null genes. Panel B illustrates that, for null genes, VST makes the variance of ${\mathcal{D}}_{{}_g}^{*}=h\left(Y_{g1}\right)-h\left(Y_{g2}\right)$ constant regardless of mean ${\mu }_{g1}$.