Fig. 2
From: MAUDE: inferring expression changes in sorting-based CRISPR screens
MAUDE correctly identifies functional elements and effect sizes in simulated data. a Simulation for mean-altering elements. A mean-altering element is modeled as changing the mean expression μg (between 0.01 and 1) of cells with the effective element (red curve) compared to cells with ineffective elements or negative control guides (gray curve). b Simulation of proportion-altering elements. A proportion-altering element is modeled as resulting in a different proportion of cells with altered expression (rg), from 1 to 100%. On average, rg% of cells would have expression increased by μg = 1 (dark red curve), and (1 − rg)% of cells would have no change in expression (μg = 0; light red curve) relative to ineffective elements and negative control guides (gray curve). c, d Effect sizes are correctly estimated. The true effect used to generate the simulated data (x axis) vs. the inferred effect size (y axis) for c mean-altering elements (where the true effect is μg) and d proportion-altering elements (where the true effect is rg). Elements meeting a 1% FDR are shown in red and others in gray. Pearson’s r is shown in lower right (considering effective elements only). Black line: y = x