Non-invasive tracking of clonal evolution and treatment response through liquid biopsies

Background: High-grade serous ovarian cancer shows high genomic instability and heterogeneous tumorsubpopulations that drive variable therapy responses and treatment failure. Tracking tumor cellpopulation changes during treatment is crucial for understanding cancer response, resistance, andmetastasis. Clonal dynamics reflects selective sweeps, fitness-enhancing genomic alterations,and spatially restricted subpopulation outgrowth. To capture these dynamics, we developed atumor-informed framework that resolves haplotype-specific copy number (HSCN) states fromsingle cell whole genome sequencing (scWGS). These high-resolution clonal profiles areintegrated into a Bayesian circulating tumor DNA (ctDNA) deconvolution method, cfClone,enabling sensitive, biologically interpretable quantification of clonal shifts from liquid biopsies. Methods: Plasma ctDNA was collected from 20 high-grade serous ovarian cancer (HGSOC) patients,along with tumor samples from multiple sites. Single-cell whole-genome sequencing (scWGS)was performed and analyzed with HapClone, a Bayesian model reconstructing the geneticstructure and dynamics of tumor clones. HapClone generated haplotype-specific copy-number(HSCN) profiles reflecting allele-specific amplifications, deletions, and structural changes. Theseprofiles were then used to deconvolute clonal contributions in plasma with cfClone, and residualanalyses revealed hidden tumor subpopulations. Results: In HGSOC cases, integrating scWGS-derived clonal states with cfDNA enabled reconstructionof tumor dynamics, quantification of clonal growth across metastatic sites, and identification ofthe dominant clone driving recurrence, indicating that metastasis is driven by the recurrent clonerather than new clones. cfClone longitudinal tracking in one case revealed tumor-fractionchanges over multiple time points, highlighting resistant subclones emerging during therapy.Including tumor content allowed detection of clones and quantitative assessment of responsedynamics, such as rapid clonal responses in another case. Differences between refractory andsensitive clones were identified, showing how longitudinal data reveal resistance mechanisms.The HSCN framework detected greater clonal diversity than state of the art structural-variant-based methods. Conclusions: scWGS-informed HSCN analysis combined with ctDNA deconvolution provides a sensitive,biologically grounded approach to track tumor evolution, identify resistant clones, and measurectDNA. This framework enables real-time monitoring of tumor changes and has strong potentialto guide precision cancer treatment by enabling non-invasive monitoring of clonal dynamics inresponse to therapeutic decisions by linking clonal dynamics to therapy decisions in a non-invasive way.