PROJECT SUMMARY Studying cancer across a diverse array of species provides a unique opportunity to interrogate factors underpinning cancer initiation and progression and facilitating the modeling of new therapeutic targets in the setting of spontaneous tumors complicated by comorbidities and metastases. While there is extensive reporting of the frequency and diversity of tumors in animals from zoos, it has not been systematically linked to the plethora of genomic resources available. To facilitate the transition of comparative oncology studies from human plus one or two other species, to pan-mammalian analyses, we propose building a pan-mammalian tumor compendium and portal to easily disseminate our resource. We will develop and apply machine learning tools to detect patterns of cancer emergence and cancer resistance in human and non-human mammalian tumors. Our approach provides opportunities to leverage the largely understudied mammalian tumor data jointly with human data to identify reciprocal links between evolution and cancer resistance. This will allow us to make new human cancer discoveries through integrative analysis of high-throughput biological data in the context of mammalian species. We offer a powerful approach combining high-quality reference genomes and genomic data from hundreds of mammalian species with machine learning, that has the promise to unearth the evolutionary genetic underpinnings that are cornerstones of cancer initiation and progression. We will develop and apply models to identify mammalian tumors that effectively mimic rare human cancers, work with collaborators to acquire tumor samples from strongest identified mammalian models, then sequence and analyze these samples to validate their effectiveness as models of human cancers.