Leveraging dog models to uncover human cancer insights

Abstract

Background: Comparative genomics can reveal insights into human disease that cannot be identified from human data alone. Dogs are particularly useful for comparative genomics studies as they share human diet and environment, and, from a genomics perspective, dogs are closely related to humans. Dogs develop spontaneous tumors that closely resemble human tumors, have a high tumor incidence rate, and their popularity as pets ensures the wide availability of samples for scientific studies. Despite these benefits, no pancancer comparative transcriptomic study of human and dog cancers exists, nor has any study systematically quantified the effectiveness of dog tumors as a model of human adult and pediatric cancers. Individual cancer studies reveal both similarities and differences between species, but the extent of the molecular similarity across tumor types remains unclear. Methods: To address this gap, we performed a pancancer analysis of 5,875 samples (913 dogs and 4,962 human samples, including 806 pediatric samples) spanning 11 tumor types. We developed a formula to quantify transcriptomic similarities between human and dog cancers across several metrics. Results: We show that, overall, dogs are an excellent model for many human cancers and that dogs tend to be better models of adult cancers than pediatric cancers, with notable exceptions such as gliomas and sarcomas. Conclusions: Our scalable approach enables rapid and accurate identification of model systems for studying human cancers, creating new opportunities for comparative oncology studies across the tree of life.