Fig. 1: Development of dependency correlation network (DCN) and Deplink analysis.

a A workflow to set up the dependency correlation network (DCN) and Deplink analysis. Step 1, acquire the dependency profiles of 17,386 genes across 1086 pan-cancer cell lines from DepMap CRISPR-Cas9 essentiality screen dataset; Step 2, for 16,854 non-essential genes, calculate the pairwise Pearson correlation score between each gene pair and generate the dependency correlation matrix; Step 3, generate DCN based on the correlation matrix using Genets; Step 4, integrate DCN with molecular profiles of pan-cancer cell lines using Deplink. b Heatmap showing cancer type-specific (top) or cancer hallmark gene set enrichment analysis (GSEA) signature-specific (bottom) dependency of DCN modules. Names of cancer types are consistent with TCGA study abbreviations. For cancer type-specific dependency analysis, the color scale shows the dependency score difference between cell lines from each specific cancer type and cell lines from other cancer types. 686 cell lines from cancer types containing at least 10 cell lines were analyzed. Only modules significantly associated with at least one specific cancer type are shown (FDR < 0.1). For hallmark signature enrichment analysis, the color scale shows the hallmark signature score difference between cell lines showing high dependency for each module and cell lines showing low dependency for that module (18% top and bottom cell lines ranked by dependency score, respectively, p-value < 0.01). Modules are ranked in the same order for both panels. c Dot plot showing DCN modules that are significantly preferentially essential to solid tumor cell lines carrying various COSMIC mutational signatures. 616 cell lines from solid tumors containing at least 10 cell lines were analyzed. d Dot plot showing DCN modules that are significantly preferentially essential to blood cancer cell lines showing high levels of various histone modifications. 70 cell lines from blood cancers containing at least 10 cell lines were analyzed. e Left, volcano plot showing the correlation between the genetic dependency of DCN modules and drug sensitivity (GDSC). The cell lines with high dependency on module #53 (CCNE1, CDK2, SKP2) are more resistant to CDK4/6 inhibitor Palbociclib. Right, the correlation between dependency and expression of genes in module #53. For b–d, p values were determined by unpaired two-tailed Student’s t-test. For e, the p-value was determined by one-way ANOVA. Source data are provided as a Source Data file.