Monitoring ctDNA to parts per million by integration of variant reads - 11th International Symposium on Circulating Nucleic Acids in Plasma and Serum (CNAPS)

Jonathan Wan ¹ Katrin Heider ¹ Davina Gale ¹ Suzanne Murphy ¹ Eyal Fisher ¹ James Morris ¹ Florent Mouliere ¹ Dineika Chandrananda ¹ Andrea Marshall ^1,3 Andrew B. Gill ² Pui Ying Chan ¹ Emily Barker ⁴ Gemma Young ⁴ Wendy N. Cooper ¹ Irena Hudecova ¹ Francesco Marass ¹ Graham R. Bignell ⁵ Constantine Alifrangis ⁵ Mark R. Middleton ⁶ Ferdia A. Gallagher ² Christine Parkinson ² Amer Durrani ² Ultan McDermott ⁵ Christopher G. Smith ¹ Charles Massie ¹ Pippa G. Corrie ² Nitzan Rosenfeld ¹

¹Cancer Research UK Cambridge Institute, University of Cambridge, Cambridge, UK
²Addenbrooke's Hospital, University of Cambridge, Cambridge, UK
³Warwick Clinical Trials Unit, University of Warwick, Cambridge, UK
⁴Cambridge Clinical Trials Unit - Cancer Theme, Addenbrooke's Hospital, Cambridge, UK
⁵Wellcome Sanger Institute, Hinxton, UK
⁶National Institute for Health Research Biomedical Research Centre, Oxford, UK

Circulating tumour DNA (ctDNA) has the potential to enhance cancer surveillance, but sensitivity limitations can inhibit its ability to detect residual disease. Patient-specific ctDNA analysis can improve sensitivity for low levels of ctDNA: by generating a large number of informative sequencing reads across multiple mutations identified by tumour sequencing, the effects of sampling error can be minimised.

Here, we outline INtegration of VAriant Reads (INVAR) which combines patient-specific analysis of hundreds to thousands of mutant loci with both custom error-suppression methods and signal enrichment approaches. We developed a bespoke background error model for patient-specific sequencing data. Furthermore, INVAR assigns greater weight to mutations more likely to be tumour-derived based on biological features identified from sequencing data.

Exome sequencing was performed on baseline tumour samples to identify a median of 625 tumour-specific mutations per patient, which were used to design hybrid-capture sequencing panels for 47 patients with stage II-IV cutaneous melanoma. We demonstrate that INVAR can detect and quantify ctDNA to parts per million, with a specificity of 96.7% in a set of independent healthy control plasma samples. We also show that INVAR may be applied to other non-targeted sequencing data.

Patients were longitudinally monitored during treatment, tracking mutant allele fractions down to 2.5ppm (1 mutant molecule per 400,000) and ctDNA levels showed a correlation of 0.8 with CT imaging volumes in patients with metastatic disease (Pearson r, P = 6.6 x 10^-10). In patients with resectable disease, post-operative detection of ctDNA was associated with a significantly shorter disease-free interval (4.5 months vs. median not reached with 5 years’ follow-up; hazard ratio 3.69, P = 0.007).

INVAR has been developed to enhance detection for ctDNA in plasma samples where tumour mutations are known from multiple sequencing data types. INVAR-like approaches may have utility for cancer monitoring as tumour sequencing becomes increasingly routine in personalised oncology.