Poster #22 - Akshaya V. Selvarajan

THE IMPACT OF CELL-FREE DNA FRAGMENTATION AND BASE-CALLING ON LIQUID BIOPSY SENSITIVITY

Akshaya V. Selvarajan (1), Subhajyoti De (1) (1) Rutgers Cancer Institute of New Jersey, Rutgers University, New Brunswick, NJ, USA

Liquid biopsy based on tumor-derived cell-free circulating DNA offers a non-invasive approach to monitor minimal residual disease (MRD), evaluate treatment response, and guide patient prognostication. Cell-free circulating DNA (cfDNA) in blood is subject to fragmentation patterns shaped by nucleosome positioning and epigenetic context, leading to non-random degradation and potential biases in sequencing coverage during liquid biopsy analysis. We quantified locus-specific coverage imbalance covering 0.4 million somatic variants from 67,129 samples representing most major cancer types, encompassing tumor and liquid biopsy sources. Our comprehensive computational analysis of the variants reported in the targeted cfDNA and tumor sequencing studies identified these locus-, allele-specific biases and other biological confounding factors that impact the sensitivity of the clinical genomic inferences. We found 440 significantly under-represented cfDNA loci covering known oncogenic hotspots within clinically actionable genes, including TP53, EGFR, and NRAS, and enrichment in functional domains that are impacted by greater degradation in cell-free DNA, leading to loss of power in the detection of low-frequency variants. Considering differences in base calling error rates, we observed allele-specific sensitivity differences at critical hotspots such as NRAS p.Q61L vs p.Q61R in the liquid biopsy samples. Longitudinal analysis demonstrated that these biases could compromise disease tracking, leading to false-negative conclusions of clonal dynamics and extinction. For example, clinically relevant EGFR and TP53 variants were missed at multiple timepoints in individual patients. Additionally, non-random cfDNA degradation impacting JAK2, IDH2, and DNMT3A hotspots may confound clonal hematopoiesis (CH) inference. Our research findings underscore the need for locus-aware, error-informed techniques to improve the reliability of liquid biopsy-based clinical management in precision medicine.