The amplification of oncogenes is one of the most common alterations found in cancer, providing cancer cells with selective growth advantages that contribute to tumour proliferation and survival. Some genome amplifications arise from double-strand breaks caused by catastrophic processes such as breakage-fusion-bridge cycles and chromothripsis. This produces circular, centromere-less extrachromosomal DNA which randomly segregates into daughter cells and, if subject to selection, can lead to gene amplification. Although oncogene amplification is known to play a significant role in cancer development, we still do not have a complete picture of the mechanisms that lead to it. Thus studying the circularization of linear chromosomal DNA can help us better understand the mechanisms behind gene amplifications. We have characterised ecDNA and other high level focal amplifications in rare cancers such as, but not limited to, ovarian carcinosarcoma, granulosa cell tumour and endometrial carcinosarcoma. We found that 35% of patients (69 out of 199) had high level focal amplifications. CCNE1, MYC and NDRG1 were the most commonly amplified genes. Of the 69 samples with high level focal amplification, to date a small fraction (~5%) are confirmed by Amplicon Architect to harbour ecDNA. This is much lower than the average frequency of ecDNA across different tumour types reported by other studies, highlighting how difficult it is to identify ecDNA from short reads alone. However we are continuing to refine these analyses. These findings will be used to identify interesting cases for generating additional data such as Nanopore long reads and methylation data. Furthermore, we plan to develop our own method to classify ecDNA that addresses some of the shortcomings of existing ecDNA classifiers while making use of the other structural and mechanistic features of ecDNA.