Castignani Viladomiu, Carla; (2025) Leveraging deconvolution approaches to study the epigenetic and transcriptomic landscape of non-small cell lung cancer. Doctoral thesis (Ph.D), UCL (University College London).

Text Castignani Viladomiu_10211660_Thesis_sigs_removed.pdf Access restricted to UCL open access staff until 1 August 2026. Download (64MB)

Abstract

In recent years, non-genetic factors and epigenetic modifications have emerged as pivotal drivers in tumour evolution. In this thesis, I have investigated the interplay between DNA methylation and genomic and transcriptomic alterations in the TRACERx study, a multi-region dataset of surgically resected lung tumours. The study of bulk methylation and expression data in cancer is hampered by cell composition and somatic copy number alterations. To address these issues, I have applied existing methylation deconvolution tools and developed a new bioinformatic method capable of disentangling tumour and normal expression profiles from bulk data. Through this work, I provide a comprehensive overview of the lung cancer methylome, revealing extensive intra-tumour methylation heterogeneity and uncovering co-occurrence patterns with genetic changes. Additionally, I expanded the identification of methylation-based driver genes and built methylation-based phylogenies that align with evolutionary histories inferred from genetic data. To study the tumour transcriptomic landscape, I developed a novel method that formalizes the relationship between allele-specific copy number, expression, and sample purity to deconvolve profiles from bulk RNA-seq. CREDAC (Copy number-based Reference-free Expression Deconvolution Analysis of Cancer) was validated using in-silico mixtures of patient-derived cell lines and scRNAseq, showing high accuracy. I revealed extensive differences in the expression program of tumour and mixed normal cells, and a consistent genomewide overexpression in cancer, with prognostic significance. I also explored expression dosage compensation upon aneuploidy, revealing a set of genes under tight regulatory control in response to copy number changes. Finally, I showed increased risk stratification using existing established prognostic signatures with CREDAC-deconvolved data compared to bulk data. In summary, by applying and developing novel deconvolution algorithms, I have contributed new insights into the functional characterization of lung cancer, with the potential to enhance our understanding of tumour biology

Download activity - last month

Export as JSONCSVXML

Download activity - last 12 months

Export as CSVXMLJSON

Downloads by country - last 12 months

Export as JSONXMLCSV

Archive Staff Only

View Item