Fig. 5
From: Computational inference of cancer-specific vulnerabilities in clinical samples
Characterization of context-dependent or synthetic vulnerabilities. a Analysis of RAN-dependent samples. Dependency of each sample (column) as estimated by the DNN model is indicated at the top. Expression levels of RAN-related genes are shown below. The triple-negative subtype was determined by each sample’s status for estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth receptor-2 (Her2). The heatmap below presents the transcription profile of the most significant 100 genes that were overexpressed in the RAN-dependent samples (P < 0.005 by the two-sample t test). Enriched functional terms are indicated on the left. b Analysis of Ku70/80-dependent samples. DNN-estimated dependency on XRCC6 or XRCC5 is marked in blue. Differential expression of the classical NHEJ genes in tumor relative to normal samples is shown. Also shown are the gene-level copy number scores of MSH3 and some genes implicated in HR (BRCA1, RAD50, and RAD51) or alternative NHEJ (LIG1 and LIG3). Negative values indicate copy number loss. P values are from Fisher’s exact test for the overrepresentation of copy loss samples in the dependent group