Toward Precision Diagnostics: Advancing Genome-wide Cell-Free DNA Profiling for Cancer and Infectious Disease Detection

Abstract

Accurate and timely diagnostics can improve treatment outcomes in both cancer and infectious diseases. When cells undergo cell death, they release short cfDNA fragments into body fluids such as blood. These fragments can originate from host cells or from pathogens, providing diagnostic information through a minimally invasive blood draw and reducing the need for invasive procedures. Genome-wide cfDNA analysis allows unbiased detection of somatic mutations, copy number alterations, fragmentomic features, and microbial DNA, thereby offering advantages over traditional targeted or culture-based diagnostic assays. Despite its promise, cfDNA-based assays are constrained by scarce signal and technical noise. In cancer, early detection and minimal residual disease monitoring require reliable detection of tumor fractions that often fall below intrinsic sequencing error rates. In infectious diseases, pathogen-derived cfDNA is typically rare and vulnerable to misclassification due to incomplete reference databases. In this thesis, I develop and evaluate genome-wide cfDNA sequencing strategies to improve detection under these constraints. The work spans both cancer and infectious disease contexts, reflecting the shared technical challenges of rare signal detection and noise suppression. This thesis develops and evaluates genome-wide cfDNA sequencing strategies to improve rare-signal detection across cancer and infectious disease settings. First, in Chapter 2, a novel technique, NanoRCS, was established as a nanopore-based consensus sequencing method that reduces random sequencing errors and enables combined detection of tumor-informed SNVs, copy number alterations, and fragmentomic features at low sequencing throughput. Then, a large-scale simulation framework in Chapter 3 defined how sequencing depth, error rate, and tumor mutational burden jointly determine the sensitivity of genome-wide SNV-based tumor detection, providing guidance for rational assay design. In Chapter 4 and 5, the thesis further applies cfDNA sequencing to pathogen detection, where microbial cfDNA is often extremely rare and vulnerable to misclassification. In Chapter 4, cfDNA is tested in a proof of principle cohort for invasive pulmonary aspergillosis detection. Optimized host genome filtering and curated fungal reference databases improved fungal detection accuracy, with simulations defining theoretical limits of detection and clinical data supporting detection of Aspergillus fumigatus. Finally, in Chapter 5, cfDNA sequencing was adapted for bacterial sepsis in newborn foals using single-stranded DNA library preparation, size selection, and contamination-aware bioinformatics. This approach revealed elevated sepsis-associated bacterial cfDNA, frequent co-elevation of multiple bacterial genera, and host cfDNA features associated with disease severity. Finally, we reflect on the findings and discuss the content of the thesis in a broader context. Looking ahead, translating cfDNA diagnostics into routine clinical practice requires attention beyond technical performance. Future efforts should focus on the clinical application of these technologies in ways that promote patient wellbeing.