Clinical sequencing efforts are uncovering numerous fusion genes in childhood solid tumors, yet few methods exist to delineate fusion-oncogenes from structural changes of unknown significance. One such novel fusion gene is VGLL2-NCOA2, which was described by us and others in patients with congenital sarcomas but has lacked functional validation. To determine if this fusion is an oncogene, and how it is driving disease, we developed a vertebrate zebrafish model and mouse allograft model of human VGLL2-NCOA2 driven sarcomagenesis. We found that VGLL2-NCOA2 is indeed an oncogene and is sufficient to generate mesenchymal tumors that recapitulate the human disease at the histological and transcriptional level. Zebrafish VGLL2-NCOA2 tumors display features of arrested skeletal muscle development, and a subset transcriptionally cluster with somitogenesis in developing embryos. By comparing tumor and embryonic gene expression signatures, we identified developmentally regulated targets that VGLL2-NCOA2 potentially leverages for tumorigenesis. These targets highlight the core biology of the disease and could represent therapeutic opportunities. Specifically, a RAS family GTPase, arf6/ARF6, involved in actin remodeling and rapid cycling of endocytic vesicles at the plasma membrane, is highly expressed during zebrafish somitogenesis and in VGLL2-NCOA2 tumors. In zebrafish tumors, arf6 protein is highly expressed and is absent from mature skeletal muscle. In VGLL2-NCOA2 mouse allograft models and patient tumors, ARF6 mRNA is overexpressed as compared to skeletal muscle or normal controls. More broadly, ARF6 is overexpressed in adult and pediatric sarcoma subtypes as compared to mature skeletal muscle. Overall, our cross-species comparative oncology approach provides evidence that VGLL2-NCOA2 is an oncogene which leverages developmental programs for tumorigenesis, and that one of these programs, the reactivation or persistence of arf6/ARF6, could represent a therapeutic opportunity.