(P 413) LINKING REAL-WORLD CLINICAL AND GENOMIC DATA IN RHABDOMYOSARCOMA: LEVERAGING AACR PROJECT GENIE AS A PATH TOWARD PRECISION ONCOLOGY IN RARE CANCERS

Objective: Rhabdomyosarcomas (RMS) are common soft tissue sarcomas in childhood, adolescence, and less frequently in adults. The WHO recognizes four histopathologic subtypes of RMS, and landmark studies have begun to identify genomic features associated with each. Still, the diverse clinical features and outcomes within this large patient population suggests that advancements in prognostic precision and treatment strategies may be improved by a deeper understanding of the genomic drivers and modifiers of disease and their associations with clinical phenotypes. Aside from FOXO1 fusion status in pediatric RMS, there are few genomic prognostic features available for risk stratification, and the clinico-genomic landscape of adult RMS, a rare cancer with poor outcomes, has not been well characterized. Utilizing Project GENIE, an international real-world clinico-genomic data-sharing consortium, we report on recurrent genomic features of RMS across the age spectrum and assess their associations with clinical phenotypes.

Methods: We analyzed targeted next-generation sequencing data of 388 tumors from 351 patients with RMS in the AACR Project GENIE registry, which aggregates, harmonizes, and links clinical-grade next-generation sequencing data with clinical data from 19 cancer institutions, each with a unique targeted panel (45-531 genes). The majority of data are from patients treated at 3 cancer centers: Memorial Sloan Kettering, Dana-Farber Cancer Institute, and University of California San Francisco. The registry currently contains 167,000 samples and is projected to reach >250,000 by 2025. We assessed the publicly available clinical and genomic features of this real-world cohort. To evaluate associations between genomic features and more detailed clinical phenotypes, we performed deeper clinical annotation on a subset of samples using PRISSMM (Pathology, Radiology/Imaging, Signs/Symptoms, Medical oncologist assessment, and tumor Markers), a novel curation framework for electronic health record abstraction at scale. Logistic regression was used to assess clinical and genomic associations, and p-values were adjusted for the false discovery rate.

Results: Fifty-seven percent of patients were < 18 years, and 43% were ≥18 years at the time of sequencing (range < 18 to 86 years). Fifty-eight percent of patients were deceased at the time of last follow-up. Sixty-eight percent of samples were primary tumors and 32% were metastatic samples. The median number of mutations per sample was 3 (range 1 to 39). Fourteen percent of patients had FOXO1 fusion positive RMS, 65% of whom were < 18 years at the time of sequencing. Amongst patients > 18 years, the mean and median age of FOXO1 fusion positive patients was 32 and 24 years, respectively (range 20-78 years), demonstrating a high prevalence even in adult RMS patients. The most frequently altered genes across the cohort were TP53, FOXO1, NF1, NRAS, BCOR, and CDKN2A. In fusion negative tumors, RAS pathway alterations were detected in >50% of cases, and cell cycle regulation pathways were altered in the remaining cases. TP53 and MDM2 alterations were found exclusively in fusion negative tumors. TP53 alterations (19%), detected at a higher frequency than previously reported, were also significantly enriched in pleomorphic RMS (q < 0.001).

Conclusion: This is the first study to genomically characterize RMS across the age continuum and to utilize publicly available real-world patient data to assess clinico-genomic associations in a large cohort of pediatric cancer patients. We demonstrate the feasibility of using such an approach to define genomic features in heterogeneous, rare cancers and highlight the value of real-world data collaboration in advancing precision oncology.