Liquid biopsies, DNA extracted from blood samples are revolutionising many areas of medicine, with an especially high impact on cancer research. In particular, low-pass whole genome-sequencing (lpWGS) of liquid biopsies makes quasi-real-time monitoring of the disease possible at minimal costs (~20€/sample). LpWGS enables the sensitive detection of copy number alterations (CNAs), large-scale genomic alterations that are highly cancer-specific, and therefore provide a quantitative insight into the amount and composition of tumour-derived DNA within the blood. However, CNA quantification becomes problematic when less than 10-15% of DNA comes from cancer cells – which is very often the case.

Here, we present a Bayesian Change Point detection algorithm applied to binned read counts to robustly identify segment (genomic region with identical CNA) breakpoints and CNA values. The derived CNA profiles are then processed using our previously developed algorithm, liquidCNA [1], a bioinformatic method to track cancer composition from CNAs measured in liquid biopsies.

We test the performance of the pipeline on synthetically generated data sets and in vitro cell line mixtures. We show that our method enables tracking tumour composition even in samples with <5% tumour content, unlocking a high number of previously unusable liquid biopsy samples.