Abstract B042: What single-tumor sequencing reveals about immune recognition and killing via neoantigen depletion curves

Abstract Immunoediting is known to constrain tumor growth, yet fundamental aspects of immune activity remain unknown. A better understanding of how tumor neoantigens are recognized and targeted by the immune system will aid immunotherapy and tumor vaccine development. Often, the available data consist of bulk sequencing from a single timepoint in tumor growth. Here we ask: what can those single time-point data tell us about when and how the immune system recognizes and then kills cancer cells? We show that tumor sequencing plus neoantigen prediction is sufficient to distinguish between major models of immune recognition and to quantify immunoediting in biologically interpretable terms. We summarize sequencing data from each tumor with a “neoantigen depletion curve”: at each cancer-cell fraction (how common a mutation is in the tumor), we compare the number of predicted neoantigenic mutations to matched non-neoantigenic mutations. In our model, once a neoantigen is recognized by the immune system, cells carrying that neoantigen are killed at a constant rate per cell. Their effective growth rate is thus negative, which reduces the prevalence of that neoantigen. Integrating over all neoantigens produced yields characteristic depletion curve shapes. We evaluate four biologically motivated models: (1) no immune recognition; (2) constant recognition rate once a given neoantigen appears in even a single cell; (3) constant recognition rate triggered only once the number of cells carrying a given neoantigen exceeds a size threshold; and (4) gradually increasing recognition rate based on the number of cells carrying a given neoantigen (as a Hill function). We fit generalized linear models to mutation counts to estimate parameters and provide diagnostics and model selection. From a single tumor, we can recover three dimensionless quantities that summarize biology without requiring growth rates from outside data: (i) the strength of immune killing relative to tumor growth; (ii) the baseline recognition rate relative to growth; and (iii) the relative neoantigen clone size of recognition (the “knee” of the curve). We highlight qualitative predictions that can be sought in data: no immune recognition yields a flat neoantigen depletion curve; constant recognition tilts the curve; recognition after a threshold creates a sharp knee with a uniform post-knee slope across neoantigen qualities; and gradually increasing recognition yields a smooth knee with stronger pre-knee depletion for higher-quality neoantigens. Together, these results show that standard tumor sequencing—without longitudinal sampling—can reveal whether immune pressure is continuous or size-triggered and can quantify its relative strength and timing. This framework offers practical estimators, uncertainty quantification, and diagnostics for immunoediting in human tumors. Citation Format: Heng Wu, Elizabeth S. Borden, Karen T. Hastings, Ryan N. Gutenkunst. What single-tumor sequencing reveals about immune recognition and killing via neoantigen depletion curves [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Cancer Evolution: The Dynamics of Progression and Persistence; 2025 Dec 4-6; Albuquerque, NM. Philadelphia (PA): AACR; Cancer Res 2025;85(23_Suppl):Abstract nr B042.