Abstract 1947: Assessment of cancer subtypes across multiple cancer types using a circulating tumor DNA (ctDNA)-based targeted methylation assay

Background: Accurate identification of cancer subtypes is essential for effective diagnosis, prognosis, and treatment selection. Traditionally, histologic type, cell of origin, and underlying pathogenesis are determined using tissue histology, requiring invasive biopsies. This study evaluates the use of a ctDNA-based targeted methylation sequencing assay to identify cancer subtypes across multiple cancer types through a tissue-free blood test. Methods: Within the Circulating Cell-free Genome Atlas (CCGA, NCT02889978) study, clinical data were recorded and plasma samples processed using the targeted DNA methylation assay. We defined a set of 14 cancer subtypes spanning histologic types (adenocarcinoma, squamous cell carcinoma, neuroendocrine cancer, urothelial carcinoma), lineages of the cell of origin (lymphoid, myeloid, plasma cell, hepatocellular, mesenchymal, melanocytic, mesothelial, Mullerian), or an oncogenic process (human papillomavirus-mediated). Pathology reports were used to assign a subtype to each sample. An “other” category was used for less frequent subtypes not captured by these subtypes. A machine learning classifier was trained using ctDNA methylation data from 3,340 non-cancer and 3,239 cancer participants with informative pathology to predict subtype. GRAIL’s Cancer Research Solution is a tissue-free, Research Use Only (RUO) multi-cancer detection test. To ensure a reliable assessment of the subtypes, cancer patients found positive by the RUO test in the held-out test set were shedding sufficient cell-free DNA and were reported for performance evaluation. Results: 1,404 out of the 2,669 cancer participants in the held-out test set were detected as cancer by the RUO test. This included 165 stage I, 293 stage II, 416 stage III, 493 stage IV and 37 non-informative stage across 54 different AJCC-defined cancer types. Classifier prediction accuracy was 92.7% (95% CI 91.3-94%, 1,302/1,404) across all 14 subtypes. In the subtypes with more than 50 samples in the data, high accuracy was observed in histologic types (e.g. adenocarcinoma: 730/749 = 97.5%), lineage of the cell of origin (e.g. Mullerian: 91/103 = 88.3%) and oncogenic process (e.g. HPV associated carcinoma: 92/96 = 95.8%). Conclusions: A ctDNA-based targeted methylation platform based on a previously validated assay demonstrated the ability to accurately predict cancer subtypes in a wide range of cancers using only a single blood draw. It offers potential in clinical decision-making, especially when tissue biopsies are unavailable, by supporting early cancer detection, identifying cancer types with unknown primaries, informing treatment selection, and monitoring resistance mechanisms (e.g. lineage switching). Further validation is needed to enhance personalized cancer diagnosis and therapy in precision medicine. Citation Format: Joerg Bredno, Rita Shaknovich, Alexis Thornton, Susanne Tilk, Oliver Venn, Chun Zhang. Assessment of cancer subtypes across multiple cancer types using a circulating tumor DNA (ctDNA)-based targeted methylation assay [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 1947.