Abstract 733: Development and characterization of a negative prediction algorithm for actionable mutations utilizing genomic and epigenomic profiling in cfDNA

Abstract Introduction: Despite success of genomic tumor characterization from plasma, one challenge of cell-free DNA (cfDNA) profiling is an inability to rule out the presence of actionable genomic mutations. This difficulty stems from the inability to assess if known drivers are absent or if the tumor level is insufficient for detection in the assay. Confident negative variant prediction could allow clinicians to expedite clinical action from cfDNA without waiting for tissue biopsy sequencing when no actionable alterations are identified. Here we leverage an assay which simultaneously profiles epigenomic and genomic signals to develop a negative prediction algorithm in advanced cancers. Methods: We use epigenomic signals to quantify an unbiased tumor fraction (TF) estimate and overcome limitations of genomic-based TF estimation including limit of detection, clonal hematopoiesis, variable copy-number on mutant alleles, and ambiguity of germline/somatic mutation origin. Using population priors, epigenomic TF, and observation of mutant and non-mutant coverage over variants of interest, we developed an algorithm to predict post-test probability of the cfDNA sample harboring guideline recommended genomic biomarkers relevant for treatment selection. This algorithm is applied in multiple tissue-specific diagnostic settings including colorectal cancer where testing for RAS/RAF hotspot mutations and lung cancer where testing multiple biomarkers are needed for optimal therapy selection. Results: In an analysis of 3775 de-identified advanced stage CRC samples, 1342 (36%) were found to have a reportable RAS/RAF mutation. Among the remaining samples, 38% had sufficient tumor fraction to confidently assess the sample as RAS/RAF variant negative (with >99% confidence), which expanded to 46% at >95% confidence. This indicates that the test can confidently assign RAS/RAF status to >65% of CRC plasma samples. Reasons why the remaining samples could not be confidently assessed as biomarker positive or negative included low tumor shedding (including 15% with nondetectable tumor), low genomic coverage over loci of interest, and mutant allele support below the confident call threshold. Finally, we compare the application of epigenomic TF against genomic tumor-fraction algorithms using either RAS hotspot or a broader subset of cancer mutations, and show the epigenomic method to be much more sensitive and precise for TF estimation than genomic-only methods. Conclusions: An epigenomics-based approach to enable confident negative prediction may increase the utility of liquid biopsy and expedite clinical decision-making in advanced solid tumors with biomarker-informed treatment pathways. Next steps include characterization in diverse tissue types as well as clinical validation to link negative call status to variant status in tissue biopsies. Citation Format: Andrew M. Gross, Katie Quinn, Leslie Bucheit, Jill Tsai, Aaron Hardin, Catalin Barbacioru, Martina Lefterova, Justin Odegaard, Darya Chudova. Development and characterization of a negative prediction algorithm for actionable mutations utilizing genomic and epigenomic profiling in cfDNA [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 733.