Double minute chromosomes (DMs) are circular fragments of extrachromosomal DNA. They are a mechanism for extreme gene amplification in some kinds of tumor cells. The existence of DMs strongly correlates with the malignant behavior of tumor cells and drug resistance. Identification of DMs and the genes they harbor is important for choosing the right therapeutic regimen for cancer treatment.

This research aims to improve an earlier framework for double minute detection (DMFinder) in next-generation sequencing data. The main contribution of this study is a procedure named AmpliconFinder, which is a Hidden-Markov Model-based approach for detecting highly amplified regions (amplicons), of which DMs are comprised. This method augments DMFinder, thus improving its robustness to noisy sequence data, and thus improving its sensitivity to detect DMs. Augmenting the original DM discovery framework with AmpliconFinder substantially improved its ability to discover DMs in simulated genomic data. Experiments on this simulated dataset have shown that AmpliconFinder helped to increase the sensitivity of DMFinder by 20% on average. Moreover, DMFinder augmented with AmpliconFinder found all previously reported DMs in three real cancer genomic datasets, whereas the original DMFinder failed to find any of them.