Figure 2 | Scientific Reports

Figure 2

From: Predicting pathogenic non-coding SVs disrupting the 3D genome in 1646 whole cancer genomes using multiple instance learning

Figure 2

Analysis of predicted pathogenic non-coding SV pairs. (a) Genes affected by pathogenic non-coding SVs as identified by svMIL2 with significant driver potential (showing top 50 most significant gene-cancer type pairs). To determine significant driver potential, random gene sets were sampled 10,000 times with the same size as the number of genes with candidate pathogenic non-coding SVs. A t-test was used to compute which of the candidate genes have more driver coding SNVs (snpEff moderate or high impact, filtered for consensus genes driven by SNVs from IntOGen) than expected by random chance. (b) Comparison of the number of genes affected by pathogenic non-coding SVs with the number of genes affected by driver SNVs reveals a preference for a different driving mechanism per cancer type.

Back to article page