Circulating free DNA multi-modal profiling to detect tissue-of-origin and direct therapy in Cancer of Unknown Primary: Taking the 'U' out of 'CUP'€™

Student thesis: Phd


Cancers of Unknown Primary (CUP) comprise a heterogeneous group of metastatic cancers where the primary tumour site is undetectable. It comprises 2% of all new cancer diagnoses; is the fifth leading cause of cancer death in the United Kingdom (UK), with few effective treatment options. The evolution of molecular profiling and tissue-of-origin classifiers has the potential to transform the management of patients with CUP, however evidence to support its use is lacking. Research in CUP is hampered by the heterogeneity of the condition and paucity of tumour tissue available for research and molecular profiling. In this thesis, I sought to overcome the challenges of tissue biopsies and evaluate the clinical utility of circulating free DNA (cfDNA) multi-modal profiling to enable better diagnosis and treatment stratification in CUP. I have evaluated the role of cfDNA mutational profiling as an adjunct or alternative to tissue biopsy in patients with CUP, revealing potentially actionable alterations in up to half of patients through copy number evaluation, mutation profiling and estimation of tumour mutation burden. Furthermore, I have contributed to the development of a bead-based methylation capture assay (T7-MBDSeq) for use with low input, cfDNA samples. This method detects tissue and tumour-specific methylation patterns from cfDNA comparable to methylation array data. I describe a tissue-of-origin classifier built from publicly available data and adapted for use on cfDNA samples, CUPiD. Applying this classifier to a validation cohort of 167 cfDNA samples from known tumour types demonstrated remarkable sensitivity (72%) and specificity (>98%) in prediction of tissue-of -origin across seven broad cancer types. Applying this classifier to a cohort of 41 patients with CUP demonstrated utility in predicting tissue-of-origin in 73% of patients, with most predictions being clinically consistent or correct predictions. Utilising mutation, copy number and DNA methylation profiling in combination revealed 78% of the cohort have either a potentially actionable alteration or tissue-of-origin that could significantly impact treatment stratification. cfDNA multi-modal profiling is feasible and showed potential clinical utility in this small cohort. 
Date of Award31 Dec 2022
Original languageEnglish
Awarding Institution
  • The University of Manchester
SupervisorElaine Kilgour (Supervisor), Caroline Dive (Supervisor) & Natalie Cook (Supervisor)


  • Molecular profiling
  • Tissue-of-origin
  • DNA methylation
  • cfDNA
  • Cancer of Unknown Primary

Cite this