Copy Number Profiling of ctDNA via Shallow WGS in NSCLC - Evaluation of an Integrated Workflow for the Detection of Somatic Alterations

Abstract

Non-small cell lung cancer (NSCLC) remains a leading cause of cancer-related mortality worldwide, necessitating robust and minimally invasive methods for longitudinal monitoring of therapeutic response and tumor evolution. Liquid biopsies utilizing cell-free DNA (cfDNA) offer a promising approach, however, capturing somatic copy number alterations (CNAs) from circulating tumor DNA (ctDNA) is highly challenging due to low tumor fractions and signal dilution from healthy background cfDNA. This thesis evaluates and optimizes an integrated workflow for copy number profiling via shallow whole genome sequencing (sWGS) at ultra-low sequencing depths (~0.0044x to 0.01x). The bioinformatic pipelines Ion Reporter, ichorCNA, and WisecondorX were systematically benchmarked using healthy controls, an undiluted A549 cell line positive control, and clinical plasma samples. To increase sensitivity for low-fraction ctDNA signals, a bisection strategy was used to adjust the hidden Markov model (HMM) parameters of an existing clinical Ion Reporter workflow. All three pipelines were evaluated across different bin sizes (0.5 Mb, 1.0 Mb, and 2.0 Mb), with ichorCNA and WisecondorX processed under default configurations. The benchmarking framework revealed distinct architectural trade-offs across the pipelines at this ultra-low depth. While the optimized Ion Reporter workflow acted conservatively to minimize false-positive calls in validation controls, WisecondorX achieved the highest overall call accuracy (Macro F1 = 0.548) on the positive control data, and ichorCNA provided the highest sensitivity in clinical samples at the expense of vulnerability to background technical noise. Ultimately, this study demonstrates that at this ultra-low sequencing depth, pipeline selection should align with the clinical objective: a conservative architecture like Ion Reporter is well-suited to reduce false positives, whereas tools like WisecondorX or ichorCNA are better suited when maximizing the detection of low-amplitude variants is the priority.