Multimodal cell-free DNA whole-genome analysis combined with TET-Assisted Pyridine Borane Sequencing is sensitive and reveals specific cancer signals

The analysis of circulating tumour DNA (ctDNA) promises to extend current tissue-specific cancer screening programmes to multi-cancer early detection and measurable disease monitoring to solid tumours using minimally invasive blood draws (liquid biopsies). Most studies so far have focussed on using targeted deep sequencing to detect the low-abundance, fragmented ctDNA. A few studies have integrated information from multiple modalities using shallow 1x WGS. Here, we developed an integrated bioinformatics pipeline for ctDNA detection based on whole genome TET-Assisted Pyridine Borane Sequencing (TAPS) of plasma samples sequenced at 80x or higher. We conducted a diagnostic accuracy study in a case-control cohort of patients presenting to the UK National Health Services (NHS) primary care pathway with non-specific symptoms of cancer, who either did not have cancer or who were subsequently diagnosed with cancer and referred to surgery with curative intent. TAPS is a base-level-resolution sequencing methodology for the detection of 5-methylcytosines and 5-hydro-methylcytosines. Unlike bisulfite-sequencing, the current established method for mapping epigenetic DNA modifications, TAPS is a non-destructive methodology, which only converts methylated cytosines and preserves DNA fragments over 10 kilobases long, thus opening the possibility of simultaneous methylome and genome analysis on the same sequencing data. The proposed methodology combines copy number aberrations and single nucleotide variants with methylation calls from TAPS-treated plasma from patients with Stage 1-4 colorectal (n=36), oesophageal (n=8), pancreatic (n=6), renal (n=5), ovarian (n=4) and breast (n=2) cancers. Plasma samples from 21 confirmed non-cancer controls were used for data denoising, while plasma samples from 9 additional agematched healthy controls were further used to establish the minimum level of detection. Copy number aberrations, single nucleotide variants, and methylation signals were independently analysed and combined in sample-specific scores, which quantify the levels of plasma ctDNA. Matched tumour samples were used for validation, not for guiding the analysis, imitating an early detection scenario. The detection threshold was set such that specificity was 100%, resulting in sensitivity of 85.2%. In silico experiments on high-fidelity synthetic data suggest excellent discriminatory capacity (AUC > 80%) at ctDNA fractions as low as 0.7%. Furthermore, we demonstrate successful tracking of tumour burden post-treatment and ctDNA shedding in precancerous adenomas in patients with colorectal cancer in the absence of a matched tumour biopsy. In summary, we developed and validated a pipeline for interrogating liquid biopsies using TAPS 80x or higher WGS that is ready for in-depth clinical evaluation both in multi-cancer screening of high-risk individuals and multi-cancer measurable disease monitoring.