CancerGD: A Resource for Identifying and Interpreting Genetic Dependencies in Cancer

Abstract

Genes whose function is selectively essential in the presence of cancer-associated genetic aberrations represent promising targets for the development of precision therapeutics. Here, we present CancerGD, a resource that integrates genotypic profiling with large-scale loss-of-function genetic screens in tumor cell lines to identify such genetic dependencies. CancerGD provides tools for searching, visualizing, and interpreting these genetic dependencies through the integration of functional interaction networks. CancerGD includes different screen types (siRNA, shRNA, CRISPR), and we describe a simple format for submitting new datasets.

Keywords: CRISPR; RNAi; cancer; genetic interactions; network biology; precision medicine; synthetic lethality.

Graphical abstract

Figure 1

CancerGD Overview Loss-of-function screens from multiple sources are integrated with exome and copy-number profiles from the GDSC resource. Cell lines are annotated according to the mutational status of a panel of driver genes (see Table S1). Statistical analysis is then performed to identify associations between the presence of driver gene alterations and sensitivity to reagents targeting specific genes. These CGDs are filtered such that only those with nominal significance (p < 0.05) and moderate common language effect sizes (≥65%) are retained. Finally, all CGDs are annotated according to whether they occur between driver-target pairs with known functional relationships (STRING) and whether there is an inhibitor available for the target gene (DGIdb).

Figure 2

Genetic Dependency Exploration and Visualization (A) The principle view of the database. Each row represents a gene identified as a dependency associated with ERBB2 amplification in Campbell et al. (2016) across all tumor types (Pan cancer). Columns display experimental details along with the p value, common language effect size, and difference in median sensitivity score for each dependency. Genes identified as dependencies in multiple datasets are indicated in the Multiple Hit column. Genes with a known functional relationship to the driver gene (e.g., PIK3CA) are indicated in the String Interaction column, and drugs known to inhibit the target gene are indicated in the Inhibitors column. Toggles/search boxes permit easy filtering of interactions, e.g., to select only those genes with an associated inhibitor available. See also Figure S1 and the tutorial in Methods S1. (B) Example boxplot showing an increased sensitivity of ERBB2-amplified cell lines to inhibition of MAP2K3. Each data point represents the sensitivity of a particular cell line to RNAi reagents targeting MAP2K3. Cell lines are grouped according to ERBB2 amplification status with the wild-type group on the left and amplified group on the right. Cell lines are colored according to site of origin and toggles on the right permit the hiding/showing of cell lines from specific sites. Hovering over a given data point provides the cell line's name, the primary site, and the score associated with the RNAi reagent in that cell line. An overlapped box-whisker plot displays the interquartile range and the median for each group. High-resolution PNG images for each boxplot can be downloaded along with a CSV file containing all of the data presented in the boxplot. Links to the target gene (MAP2K3) on external websites are provided at the bottom of the plot.