Abstract
Interpreting the function of noncoding mutations in cancer genomes remains a major challenge. Here we developed a computational framework to identify risk noncoding mutations of all classes by joint analysis of mutation and gene expression data. We identified thousands of SNVs/small indels and structural variants as candidate risk mutations in five major pediatric cancers. We experimentally validated the oncogenic role of CHD4 overexpression via enhancer hijacking in B-ALL. We observed a general exclusivity of coding and noncoding mutations affecting the same genes and pathways. We showed that integrated mutation signatures can help define novel patient subtypes with different clinical outcomes. Our study introduces a general strategy to systematically identify and characterize the full spectrum of noncoding mutations in cancers.
Footnotes
Disclosure of potential conflicts of interest: The authors declare no competing interests.
