Development and clinical validation of a multi-modal liquid biopsy test for cancer screening and monitoring using long-read sequencing technologies

Background.

Detecting cancer early is fundamental to improve survival rates, as prognosis for advanced and metastatic cancers is dismal.

Liquid biopsy (LB) methods based on detecting tumour-specific mutations in circulating tumour DNA (ctDNA) represent a promising strategy for early detection.

Nevertheless, current LB methods have low sensitivity, low specificity, and require high upfront investment in sequencers and sophisticated experiments, all of which hampers their wider impact in clinical settings. Overall, developing cost-effective technologies for early detection represents a major unmet clinical need.

Nanopore sequencing is an emerging technology that permits continuous reading of individual DNA molecules in real time without requiring preamplification steps, thus enabling fast turnaround times and concomitant detection of copy number, structural variants, small mutations, and methylation profiles in ctDNA in the same assay.

Although several studies have shown the feasibility of long-read sequencing for disease monitoring, its potential for early detection remains largely unexplored. Aims.

I aim to develop and validate a highly sensitive, highly specific, multi-modal, cheap and fast turnaround plasma-based LB test.

This test will be fundamentally different from existing ctDNA analyses in that it is based on an entirely new sequencing technology and data analysis approach that will address the sensitivity and specificity limitations of LB described above, while also improving turnaround times, portability and costs.

Methods.

The data analysis component of the project will rely on successful preliminary results that have established Artificial Intelligence methods for the detection of mutations in long-read sequencing data.

To improve sensitivity, I will use adaptive sampling, combined with PCR-barcoded amplification of ctDNA and Cas9-based amplification-free enrichment methods.

To increase specificity, I will develop algorithms to detect in ctDNA the mutation types described above, which will be combined using machine learning to predict which patients have cancer.

I will validate the test prospectively for cancer screening (Lynch syndrome patients) and disease monitoring (high-risk pediatric cancer) using blood samples collected over years before (pre)cancer diagnosis.

Focusing on high-risk patient populations whose tumours reflect the molecular diversity of tumours in low-risk populations will reduce sample size and costs to estimate the sensitivity and specificity of the test at high precision. How the results of this research will be used.

By improving early cancer detection using a minimally invasive test, I aim to make significant impact to prevent thousands of cancer-related deaths per year, and to ultimately make early detection in clinical practice a routine activity.