Non-Invasive cancer detection: computational applications in liquid biopsy and radiomics

This thesis explores the field of non-invasive cancer detection, in particular, it focuses on the development and application of computational approaches for analyzing liquid biopsy and radiomic data. The research presented herein addresses the limitations of current methods in comprehensively analyzing circulating tumor DNA (ctDNA), particularly at low ctDNA fractions, and introduces novel tools and strategies to enhance the sensitivity and specificity of ctDNA analysis. The first part of the thesis focuses on the development of Synggen, a tool for generating synthetic cancer data. Synggen addresses the critical need for standardized datasets to benchmark cancer analysis tools by creating synthetic data that closely mimics real-world samples. This tool allows researchers to evaluate algorithms in a controlled environment, overcoming the limitations associated with relying solely on real patient data, such as privacy concerns and inherent heterogeneity. The second part of the thesis introduces eSENSES, a novel NGS panel designed for sensitive and specific ctDNA characterization in breast cancer through the analysis of somatic copy number alterations (SCNAs). The panel's unique combination of elements, including genome-wide and focal single nucleotide polymorphisms (SNPs), exonic regions of key cancer-related genes, and a bespoke computational strategy, enhances the detection and quantification of SCNAs and ctDNA fraction. This approach aims to improve disease monitoring and inform therapeutic decisions in breast cancer patients. The final part of the thesis investigates the potential of radiomics in predicting treatment response in lung cancer patients. This exploration aims to complement liquid biopsy approaches by incorporating quantitative features extracted from medical images.