Perturbations in TSG pairs promote tumorigenesis in vivo. A, Schematic of combinatorial CRISPR screening in vivo. B, Cumulative distributions of dual sgRNA abundance in the plasmid library and tumors. *, P < 0.05; Kolmogorov–Smirnov test. C, Percentage of dual sgRNAs with log2 RPM ≥1 in the PTEN−/−, PIK3CA, and MYC cell libraries and tumors. Data are presented as mean ± SEM. Mean values were calculated from nine PTEN−/−, three PIK3CA, and six MYC tumors. *, P < 0.01; one-sided Student t test. D, Relative percentage of DKO/SKO normalized counts in the PTEN−/−, PIK3CA, and MYC cell libraries and tumors. *, P < 0.01; one-sided Student t test. E, Heatmap of Euclidean distances between all the cell and tumor samples (PTEN−/−). F, The effect size (LFC) of DKOs and SKOs in the PTEN−/− context. A total of 71 oncogenic perturbations are highlighted in red. G, The top 10 tumor-promoting DKOs and corresponding SKOs in the PTEN−/− context. The mean is plotted along with the 95% CI. H, Heatmap comparing the single-gene tumorigenic effects in the PTEN−/−, PIK3CA, and MYC contexts. I, Scatter plot of average log2 RPM in the PTEN−/− tumors versus preinjected cell libraries. A total of 71 oncogenic perturbations shown in F are highlighted in red, and the top 6 ones are labeled. The linear regression line is shown (dashed). J, Scatter plot and correlation analysis of the tumorigenic effects measured by quantile analysis versus average based regression (PTEN−/−).