A novel active chromatin cell-free DNA (cfDNAac) assay for early detection of colorectal cancer.

Abstract

3605 Background: Colorectal cancer (CRC) is the third most common cancer and the second leading cause of cancer-related deaths globally (Xi et al. 2021). In the U.S., it is estimated that 152,810 new cases and 53,010 deaths will occur in 2024, with an increasing incidence among younger adults (ACS, 2024). CRC is a multifactorial disease influenced by both genetic and environmental factors and typically progresses from normal epithelial tissue to adenocarcinoma through a well-established sequence of molecular events. Despite advancements in treatment, early detection remains a key driver of improved survival outcomes. While current screening methods— such as fecal immunochemical testing, fecal DNA testing, and colonoscopy— have reduced CRC mortality and morbidity, low compliance continues to be a significant challenge. In this study, we developed a proof-of-concept machine learning (ML) classifier leveraging active chromatin cell-free DNA (cfDNA ac ) signals in peripheral blood to differentiate CRC patients from healthy individuals. Methods: Plasma samples from treatment-naive colorectal cancer (CRC) patients (stages I–IV, n = 54), advanced adenoma patients (n = 14), and healthy volunteers (n = 40) were processed using Aqtual’s proprietary active chromatin capture workflow to enrich regulatory-active chromatin cfDNA. Samples were split into a training set, which included 31 early-stage CRC samples (stage I: n = 13, stage II: n = 13, stage III: n = 5) and 19 healthy samples, and a hold-out set, comprised 37 disease samples (stage I/II: n = 6, stage III: n = 6, stage IV: n = 11, advanced adenoma: n = 14) and 21 healthy samples. A machine learning classifier was trained using 5-fold cross-validation with 5 repeats on the training set, and the performance was assessed on the hold-out set to ensure robustness and generalizability. Results: A machine learning classifier trained on genome-wide cfDNA ac signals demonstrated robust performance, achieving a mean AUC of 0.94 (95% CI: 0.92 - 0.96). It exhibited 93% sensitivity (95% CI: 84% -100%) for colorectal cancer detection (Stage I/II: 94%, Stage III: 95%, Stage IV: 91%) and 81% sensitivity (95% CI: 67% - 95%) for advanced adenoma, with a specificity of 90% in the hold-out set. Analysis of the top contributing cfDNA ac signals identified 852 promoter and 2,594 exon features derived from 2,678 unique genes. Gene-set enrichment analysis revealed significant associations with key cancer hallmarks, including KRAS signaling and epithelial-mesenchymal transition (EMT). Conclusions: This study highlights the potential of Aqtual’s active chromatin capture assay to identify molecular features in plasma that differentiate CRC and advanced adenoma from healthy individuals. The ML classifier based on these signatures shows promise for future development as a non-invasive tool for early colorectal cancer and advanced adenoma detection and monitoring.