iDEG is robust to its the assumptions. Panel A indicates robustness of iDEG to the assumption that \(\delta_{g}\) is a function of expression mean \(\mu_{g}\). The \(F_{1}\) scores of iDEG are the highest among the four methods, when the values of \(\delta_{g}\) are randomly drawn from a uniform distribution \(\text{Uniform}(0.001,0.1)\). Panel B indicates robustness of iDEG that the majority of the genes are null genes. In this panel, the \(F_{1}\) scores of edgeR approach the scores of iDEG at unrealistically high percentages of DEGs.