Fig. 4

Independent validation of model performance using a saturating screen of purine biosynthesis genes. A Transcriptional architecture of the 9 targeted purine biosynthesis genes in E. coli K12 MG1655. All possible gRNAs were designed for each gene; each blue vertical line represents a gRNA. In total, 750 gRNAs were designed. Grey boxes represent genes, black arrows transcriptional start sites. B Spearman correlations between the predicted scores and measured logFC across collected timepoints. C Enrichment of efficient guides, calculated as the percentage of the experimentally determined 20% most efficient gRNAs in the predicted 20% most efficient gRNAs. The gRNAs are ranked within each gene. D Positive predictive values for all gRNAs across all time points. The predicted positives are defined as the top 3 predicted most efficient gRNAs for each gene, while the positive class includes gRNAs within N-fold of the depletion value of the most strongly depleted gRNA for each gene (N = 1.5–5 with a step size of 0.5). The boundaries of the shaded regions and points indicate PPV values for each time point. The genes purE and purK were excluded in B–D. E Measured logFCs for each guide as a function of distance to the start codon for purC, purD, and purE. The shaded regions indicate the 95% confidence interval for the fitted regression line. Plots for the other 6 genes included in the screen are given in Additional file 1: Fig. S7E