Figure 1. Integration of drug and CRISPR gene dependencies in cancer cell lines.

1. Linear models were used to integrate drug sensitivity (IC₅₀ values) and gene fitness measurements.
2. Volcano plot showing the effect sizes and the P‐value for statistically significant associations, Benjamini–Hochberg false discovery rate (FDR)‐adjusted likelihood‐ratio test P‐value < 10%. Drug–gene associated pairs are coloured according to their shortest distance in a protein–protein interaction network of the gene to any of the nominal target of the drug.
3. Percentage of the 358 drugs with significant associations and their shortest distance in the PPI network to the drug nominal targets. T represents drugs that have a significant association with at least one of their canonical targets, “−” represents no link was found, and X are those which have no significant association.
4. Examples of the top drug response correlations with target gene fitness. Each point represents an individual cell line. MCL1_1284 and venetoclax are MCL1 and BCL2 selective inhibitors, respectively. Gene fitness log₂ fold changes (FC) are scaled by using previously defined sets of essential (median scaled log₂ FC = −1) and non‐essential (median scaled log₂ FC = 0) genes. Drug response IC₅₀ measurements are represented using the natural log (ln IC₅₀).
5. Kinobead affinity is significantly higher (lower pK _d) for compounds with a significant association with their target (n = 20, Mann–Whitney P‐value = 3.1e‐07). Box‐and‐whisker plots show 1.5× interquartile ranges and 5–95^th percentiles, centres indicate medians.