Mapping DNA replication using nanopore sequencing

Magali Hennion from the Institute of Biology of the École Normale Supérieure in Paris began by explaining how DNA replication happens and how its dysregulation causes DNA damage and genomic instability. Dysregulation of DNA replication is important in tumorigenesis.  Magali explained there are currently genome wide population level methods to study DNA replication but with these methods replication fork speed and cell to cell variability cannot be measured. There are methods that can measure at the single molecule level (i.e FISH) but these are time consuming, low throughput and lack resolution. To achieve a genome wide single molecule approach they looked to nanopore sequencing using the thymidine analogue BrdU to mark actively replicating regions of the genome. Comparing cells grown on BrdU to normal cells, BrdU sites showed clear shifts in current levels distinguishable from the thymine canonical base.

Using a pulse chase experiment in asynchronous Saccharomyces cerevisiae (MCM869) cells Magali was able to use two different computational methods for identifying BrdU incorporation one using Recurrent Neural Networks (RNN), trained on datasets including BrDU instead of thymidine and the other using Transitional Matrices (TM) based on differential current shift of T to BrdU. Both methods enabled detection and orientation of >30,000 individual replication tracts, showing good concordance between the RNN and TM method. Replication Fork directionality (RFD) profiles were generated from the data, the vast majority coincide with the well characterised ARSs (known origins). The RFD profiles also correlated well with short read Okazaki fragment sequencing. Using this single molecule method Magali saw 4% of initiation events were observed outside known origin sites, suggesting there are low frequency origins that cannot be detected with population levels

Magali finished by describing this method as a reliable way to detect and orientation at high resolution and high sensitivity much quicker than existing methods.