Abstract 5936: Landscape of epigenomic tumor fraction in large pan-cancer cfDNA cohort

Abstract Introduction: Cell-free DNA (cfDNA) tumor fraction (TF) quantification is an emerging biomarker with many promising applications, including the potential to complement tumor size measurements from imaging. Genomic approaches for TF estimation are limited by the presence of cancer driver alterations, detection sensitivity, and may be confounded by non-tumor derived variants. Therefore, we have previously developed a method using differentially methylated regions to robustly estimate TF (Chen et al AACR 2023). Here we leverage a rich dataset of >50, 000 cfDNA samples from >20 cancer types to better understand the variability of epigenomic signals in cancer and non-cancer individuals and inform an update of this TF estimation algorithm. Methods: We employed a data-driven approach to update the TF algorithm leveraging a large cohort of cfDNA samples with simultaneous genomic and epigenomic profiling. This extensive dataset allowed for robust modeling of non-tumor variability in epigenetic profiles, and better calibration of tumor signals across diverse primary tissue sources shedding signals into plasma. The method update includes a data-informed region selection targeting high pan-cancer signal to noise, as well as more robust accounting for rare sources of non-tumor methylation signals including technical variation and treatment associated tissue damage. Results: We evaluated the performance of epigenomic TF estimates using 1203 samples with somatic maximum MAF, establishing 88% concordance within 0.2-5X range. Next, samples were computationally titrated into non-tumor background samples, demonstrating consistent concordance with genomic data at up to 10-fold dilutions and maintenance of precision well below the resolution of genomic profiling, largely driven by algorithmic improvements in this updated release. Case studies of discordant samples show the benefit of TF over genomic quantification, especially in scenarios with an ambiguous genomic driver, or low genomic precision. In analytical validation studies of experimental titrations over replicates, linearity and quantitative precision of TF was demonstrated down to an order of ∼0.01%, compared to >0.4% for genomic TF. Conclusions: Our liquid only TF has been analytically validated for use in clinical, research, and trial settings in patients with solid tumors. In addition to its use for absolute quantification, it is incorporated into downstream applications including therapy response monitoring and algorithms for labeling germline variants, clonal hematopoiesis, and wild-type variant calling. Finally, its application in a clinical cohort of over 50, 000 samples has led to insights into the dynamics of cfDNA tumor shedding across pre- and on- treatment individuals improving accuracy of quantification and providing results for more patients. Citation Format: Andrew Gross, Jun Zhao, Katie Quinn, Sai Chen, Kyle Chang, Catalin Barbacioru, Tingting Jiang, Carin Espenschied, Kristin Price, Jack Tung, Darya Chudova. Landscape of epigenomic tumor fraction in large pan-cancer cfDNA cohort [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 5936.