Figure 2.

DEGs detected across linear models and software. AUC was calculated from curves of 1 − specificity to sensitivity for P-value cutoffs from 0 to 1 in 100 simulated datasets. When available, models were corrected for gene-level weights (kimma, limma, dream) and/or paired design (kimma, limma, dream, DESeq2, edgeR) as indicated by panel headers. Black squares indicate means with standard deviation bars. Across all panels, letters indicate statistically different groups, Tukey FDR < 0.05, in order from highest AUC (a) to lowest (g). Points are colored by software package.