Extended Data Fig. 2: Validation of Mixscale perturbation scores across different sgRNAs and datasets.
(a) Scatter plots illustrating the relationship between the expression level of the perturbation targets (y axis) and the perturbation scores (x axis) in each cell. This plot is analogous to Fig. 2b but this time cells are stratified by their guide RNA identities instead. (b) Single-cell heatmap for STAT1 perturbation in three cell lines after IFNγ stimulation, split by sgRNA identities. (c) Comparison of Mixscale score and target gene expression estimated in an external CRISPRi dataset (Jost et al 2020 Nat. Biotechnol.). The figure displays the Mixscale score (y axis on the left) using black dots and the degree of knockdown of the target gene (y axis on the right) marked by red triangles. The x axis represents different sgRNAs, including the perfectly matched sgRNA (“_00”) and those with varying numbers of mismatched nucleotides. Plot shows that sgRNA that result in more effective knockdown also result in cells with higher Mixscale scores. (d) Comparison of Mixscale score and relative activity of sgRNAs. Similar plot as in (c), but instead the figure contrasts the Mixscale score (black dots) with the relative activity of the sgRNA (y axis on the right) marked by blue diamonds, a phenotypic measure of cellular growth defects measured from a viability screen. Plot shows that sgRNA with the highest phenotypic activity also yield cells with the highest Mixscale scores. Source numerical data are available in Source Data.
