@phdthesis{01KWG3WD02D7FJ9BXJE7R0N71M,
  abstract     = {{Accurate tumor classification is essential for guiding clinical management, yet remains
challenging in cases with limited, degraded, or highly heterogeneous DNA material. This
dissertation investigates the potential of a novel technique called cfRRBS, developed at the
UGent-VIB, as a cost-efficient and low-input method for DNA methylation profiling. The
overarching goal is to evaluate whether cfRRBS can support reliable cancer diagnosis in both
formalin-fixed paraffin-embedded (FFPE) tissue and liquid biopsies, with a particular focus on
cancers of unknown primary (CUP) and central nervous system (CNS) tumors.
The first part of this dissertation focuses on optimizing cfRRBS for cancers of unknown primary
(CUP) by establishing standardized sample workflows for plasma, ascites, and pleural effusion
samples, followed by assessment of circulating cell-free DNA (cfDNA) quality and integrity. To
enable cfRRBS-based tumor classification, existing public resources such as The Cancer
Genome Atlas (TCGA) were found to have limited CpG site overlap with cfRRBS. Consequently,
the development of a custom cfRRBS-based reference dataset covering 16 tumor entities and
healthy plasma samples became a central component of this work. Entity-specific methylation
regions were defined for each entity to build a classifier based on nonnegative least squares
deconvolution.
Methylation profiling on FFPE-derived DNA is already routinely performed in clinical practice for
CNS tumor diagnostics. The second part of this dissertation addresses the benchmarking of
cfRRBS against the current clinical gold standard—the Illumina Infinium MethylationEPIC array
and its associated Heidelberg classifier for CNS tumor classification. To enable compatibility
with array-based references, a cfRRBS-specific classifier for CNS tumors was constructed using
a classification strategy adapted from the CUP workflow, using only CpG clusters shared
between platforms. Sensitivity was assessed through in silico dilutions and reproducibility
through inter- and intrarun comparisons.
Across both applications, cfRRBS showed strong potential for accurate tumor classification,
even from fragmented or low-quantity DNA. This work highlights cfRRBS as a versatile and
accessible technique for methylation-based cancer diagnostics on both FFPE tissue and liquid
biopsies.}},
  author       = {{De Wilde, Jilke}},
  language     = {{eng}},
  pages        = {{217}},
  publisher    = {{Ghent University. Faculty of Medicine and Health Sciences}},
  school       = {{Ghent University}},
  title        = {{Advancing cancer diagnostics through DNA methylation profiling of liquid and FFPE biopsies}},
  year         = {{2026}},
}