Influence of genetic ancestry on UV mutational signatures linked to immunotherapy response in melanoma, beyond TMB.

Abstract

3087 Background: Melanoma is a UV-driven malignancy with substantial biological heterogeneity, yet genomic profiling efforts have historically underrepresented diverse ancestral backgrounds, limiting insights into population-level differences in tumor biology and treatment response. Methods: In this retrospective study, a supervised machine learning framework was developed to infer genetic ancestry across 483,454 tumors profiled by tumor-only whole-exome sequencing (WES) at Caris Life Sciences (Phoenix, AZ), using 36,962 ancestry-informative variants derived from globally diverse Genome Aggregation Database (gnomAD) reference individuals (n=4,150). Survival outcomes inferred from claims data were evaluated using Cox proportional hazard ratios (HR) with 95% confidence intervals (CI) and log-rank p-values. Associations between features were evaluated using standard statistical tests and regression models. Results: We applied this framework to cutaneous melanoma (N=8,872). The UVB-related COSMIC mutational signatures SBS7a and SBS7b were markedly enriched in tumors from patients (pts) of European (EUR) ancestry (93.2% and 75.6%, respectively) but were uncommon in African ancestry tumors (6.9% and 5.2%). Within EUR ancestry pts, higher EUR ancestry fraction was associated with higher SBS7a and SBS7b prevalence (p=9.0x10 -4 and 1.5x10 -4 ), consistent with known differences in pigmentation biology and melanoma etiology. Among 3,587 melanoma pts treated with first-line immune checkpoint inhibitors (ICI), SBS7a positivity was strongly associated with improved overall survival (OS HR [95% CI] = 0.55 [0.46–0.66]) and time to next treatment (TTNT HR=0.58 [0.49–0.69]; all p<0.001), outperforming tumor mutational burden (TMB) as a prognostic biomarker. SBS7a+ remained prognostic in a multivariable Cox analysis accounting for TMB, age, sex, and the first two ancestry principal components. Furthermore, both SBS7a+ TMB-high and SBS7a+ TMB-low tumors demonstrated superior outcomes compared to SBS7a‒ TMB-low tumors (OS HR=0.49 [0.41–0.60] and 0.62 [0.51–0.74], respectively; all p<0.001). Similar patterns were observed in cutaneous squamous cell carcinoma. To assess specificity for UV-driven biology, we evaluated two non-sun-exposed malignancies: uveal melanoma and renal cell carcinoma (RCC). SBS7a was virtually absent in uveal melanoma and RCC (6.7% and 6.9%, respectively, among EUR ancestry pts) and did not stratify ICI outcomes in either cancer. These negative controls suggest that SBS7a does not arise without UV exposure and is not a general immunotherapy biomarker. Conclusions: These findings establish a scalable genetic ancestry framework, reveal ancestry-linked differences in melanoma mutagenesis, and demonstrate that SBS7a provides clinically meaningful, TMB-independent prognostic information specific to UV-driven tumors.