Abstract 10: A continuous modeling framework for age-dependent selection on melanoma driver mutations.

Abstract Aging is a major risk factor for melanoma, yet whether age alters tumor evolution by altering mutation rates, the strength of selection on drivers—or both—has not been investigated. We developed a continuous age-selection modeling framework that estimates strength of selection as a function of patient age from large-scale sequencing data. Using whole-exome (TCGA) and targeted-panel (AACR Project GENIE) data from 5,472 melanomas, we inferred neutral mutation rates from synonymous substitution counts and mutational covariate data within <40, 40-59, and ≥60 year age groups, calibrating site-specific rates within each tumor using COSMIC mutational signatures. Age-dependent selection on individual variants was then modeled within a shared Poisson-likelihood framework using three parametric forms—linear, logistic, and generalized sigmoid—allowing flexible continuous relationships between age and the strength of selection. Model fits were evaluated by likelihood-ratio tests. To quantify uncertainty, we used adaptive Metropolis-Hastings sampling around the maximum-likelihood estimates to build likelihood-ratio-based confidence intervals and to estimate the posterior probability that the inferred age effect reverses direction. Applied to melanoma, neutral mutation rates rose modestly with age, and UV-associated signatures dominated at all ages, but endogenous processes (SBS1, SBS3) increased in older tumors. In contrast, this framework revealed strongly age-dependent and variant-specific selection: BRAF V600E exhibited a steep, continuous decline in selection with age; NRAS Q61 substitutions showed the opposite pattern, with increasing selection in older patients; and TERT-promoter mutations (C228T, C250T) became more frequent yet conferred weaker selective advantage with age. Gene-level analyses of TP53, PTEN, and CDKN2A revealed monotonic declines in selection, consistent with convergently lower oncogenicity in late-life tumors. At BRAF codon 600, single- and double-base substitutions (DBS) producing V600E provided similar selective benefits, despite the rarity of DBS events, whereas the DBS variant V600K showed increasing selection with age. Together, these results demonstrate that age shapes the somatic evolution of melanoma primarily through age-dependent selection rather than mutation supply or identity, mechanistically tying the differing burdens of BRAF- versus NRAS-driven disease across the lifespan to the differential state of aging tissues and providing a framework for age-aware cancer prevention, molecular testing, and therapeutic strategies. Citation Format: Yihan Liu, Moein Rajaei, Jeffrey D. Mandell, Jorge Alfaro-Murillo, Jeffrey P. Townsend. A continuous modeling framework for age-dependent selection on melanoma driver mutations [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 10.