Fig. 2

In vitro perturb-seq against oncogenic drivers and modifiers of radiotherapy response in GBM cells. A Schematic of in vitro perturb-seq workflow against target genes nominated by genome-scale CRISPRi screens, combined with treatment with or without 2 Gy × 5 fractions of radiotherapy. B Distribution of mean on-target knockdown levels for all in vitro perturbations in either treatment condition. Box plots show 1st quartile, median, and 3rd quartile; whiskers represent 1.5 inter-quartile range. p values indicate Mann–Whitney U test compared to non-targeting controls. C Bubble plot of gene set enrichment analyses of gene expression modules (rows) following sgRNA perturbations (columns) without radiotherapy. Expression represents log₂ fold change normalized to non-targeting controls. Top bar charts show mRNA remaining of target gene, growth (gamma), radiation (tau), and radiation to growth ratio (rho) in vitro screen phenotypes. Gene ontology of perturbed target genes indicated. D As in C but in radiotherapy conditions. Perturbation phenotypes were normalized to irradiated cells expressing non-targeting sgRNAs. E LDA UMAP plots showing distribution of single cells expressing the indicated sgRNAs in either no radiotherapy (top) or radiotherapy (bottom) conditions. Cells which express sgRNAs other than the one highlighted in color are indicated in gray. F Difference in UMAP gaussian kernel densities between perturbations in radiotherapy (RT) and no treatment (no RT) conditions. Each point represents a particular genetic perturbation. Gray bar = mean