Fig. 3

Inferring prevalence of transcriptional adaptation at single-cell resolution in Perturb-seq data. A Schematic of experimental design and paralog gene expression analysis, adapted from Frangieh et al., 2021. 143 gene targets, with a consistent batch of non-template control-treated cells, passing quality control filters (see the “Methods” section). B Perturb-seq-based single-cell paralog expression change after reference gene knockout. Per paralog, percentage of cells positive for that paralog’s expression in non-template control guide-treated cells on x-axis, percentage of cells positive for that paralog’s expression in cells treated with guides targeting a reference CRISPR target for which the gene is a paralog. For paralog genes with percent-positive > 75% in non-template controls, mean expression plotted in inset at right. Paralogs ranked in the top-100 of absolute increases per quantification method marked in magenta. All paralogs of all knockouts shown meeting minimum cell count and UMI count in gray, inclusion criteria listed in the “Methods” section. C Venn diagram summarizing the number of paralog-CRISPR target pairs with the top-100 largest increases in mean expression and/or percent-positive expression levels. 1792 total paralog-target pairs, of which 1677 are not in the top-100 largest increase lists. 85 paralog-target pairs in the top-100 largest difference list by both mean and percent-positive analysis