Fig. 2

Enhancement of CRISPR screen result. a For K562 CRISPRi screen under Cholera/Diptheria toxin selection, the gold standard of toxin protective genes comes from KEGG pathways ‘Vibrio Cholera Infection’ and ‘Glycosphingolipid biosynthesis’. For each gene, NEST computes the neighbor CRISPR score as the sum of CRISPR screen fold change scores of neighbor genes connected in the network. The prediction performance is compared between NEST and original CRISPR scores. b The gold standard of toxin sensitizing genes comes from KEGG pathways ‘Ribosome’ and ‘Proteasome’. The prediction performances of NEST and original CRISPR values are compared. c The original CRISPR values were randomized by Gaussian white noise. The standard deviation of all original CRISPR values was used as base level. At each noise level relative to the base level, the area under ROC curve (AUC) of prediction is compared with the initial AUC for toxin protective genes in K562 Cholera toxin screen. The reduction ratios were plotted for NEST and original CRISPR scores. d The reduction ratios were plotted for toxin sensitizing genes. e The STRING network was randomized and the NEST scores were calculated for 1,000 random networks. We used multivariate logistic regression to test the association of NEST scores with gold standards, after controlling the effects from network degree and original CRISPR score. The Logit Z-scores are shown for real and random networks. f As an example of high NEST score gene, COG6 is a component of Golgi complex, and connected with several other components in Golgi complex. The thickness of each edge represents the interaction weight. The color of each gene represents the CRISPR screen fold change score; red color indicates toxin protective and blue color indicates toxin sensitizing