Automated analysis of DNA replication fork progression in the human genome with ForkML

Replication fork progression is a critical determinant of genome stability, cell fate, and embryonic development. However, traditional approaches to measure fork speed, such as DNA combing or spreading, remain limited by a low throughput and a high technical complexity. To overcome these barriers, we present ForkML, a high-throughput, automated pipeline for measuring replication fork velocity in the human genome. The progression of DNA replication forks is tracked in vivo by double pulse-labelling of asynchronously growing cells with BrdU, which is subsequently digitised using Oxford Nanopore sequencing. BrdU incorporation is directly detected on long nanopore reads using custom nanopore basecalling models, and analysed through our machine learning framework trained on curated annotations. Fork directionality is inferred from the characteristic patterns of BrdU incorporation, which competes with endogenous thymidine during DNA synthesis. Fork speed is estimated simply by measuring the distance between the starts of codirectional BrdU-labelled tracks, divided by the time interval between pulses. Applied to HCT116 and HeLa cells, ForkML identifies thousands of forks per run, reliably recapitulates known fork speeds across replicates, and sensitively detects replication stress-induced fork slowing caused by hydroxyurea or aphidicolin treatments. Importantly, direct fork mapping by ForkML connects replication dynamics to genomic context, revealing significant variations in fork velocity related to replication timing, chromatin state, and transcriptional activity, exposing a slower fork progression in early-replicating, highly transcribed regions. ForkML provides a scalable, automated, and high-resolution alternative to DNA fibre assays. Compatible with standard Oxford Nanopore workflows and adaptable to various cell types, it provides a powerful new tool to study genome replication under normal and stress conditions.