Mapping the genomic landscape of DNA damage with mutli-level decoding of nanopore sequencing data

Nanopore-based direct sequencing of nucleic acids presents the opportunity to simultaneously detect, map, and strand-phase DNA modifications and damage. The signal is in the raw sequencing data but the identification of chemically modified DNA depends on computational models trained to recognise each modification in diverse sequence contexts. We have developed an experimental approach chained-duplex sequencing (CD-seq) specifically for the efficient production of training data. This project aims to utilise the abundant training data to build high accuracy machine learning (ML) models for DNA damage identification and mapping. The models will be applied to produce genomic maps of DNA damage and repair that are important for research but have translational potential in surveilling exposure to environmental genotoxins and optimising chemotherapeutic treatments of cancer.