Identification of co-occurring insertions in cancer genomes using association analysis

Collections of tumor genomes created by insertional mutagenesis experiments, e.g., the Retroviral Tagged Cancer Gene Database, can be analyzed to find connections between mutations of specific genes and cancer. Such connections are found by identifying the locations of insertions or groups of insertions that frequently occur in the collection of tumor genomes. Recent work has employed a kernel density approach to find such commonly occurring insertions or co-occurring pairs of insertions. Unfortunately, this approach is extremely compute intensive for pairs of insertions, and even more intractable for triples, etc. We present a novel approach that combines kernel density and association analysis (frequent pattern mining) techniques to efficiently find commonly co-occurring sets of insertions of any length. More generally, this approach can be used to find other commonly occurring features in collections of genomes.