Fig. 11

Effect of positive phenotypes on growth-based screens. For growth-based screens, the presence of genes with positive phenotypes (fitter than wild type) strongly influences hit detection as a function of screen duration. Screens were simulated for a set of genes in which 10% of all genes had negative phenotypes (less fit than wild type), and 2% of genes had positive phenotypes. The strength of positive phenotypes was varied, as encoded by the heat map. Hit detection was quantified separately for genes with negative phenotypes (top row) and genes with positive phenotypes (bottom row). Simulations were carried out for screens with different durations, as measured by the number of passages. Lines and light margins represent means and 95% confidence intervals, respectively, for 25 independent simulation runs. In a and c, hit detection is measured as Area under the Precision-Recall curve (AUPRC), as detailed in the Implementation section