Oxford Nanopore releases new machine learning model to enable real-time, high-accuracy epigenetic insights in biological analyses

Oxford Nanopore expands capabilities for methylation analysis, giving users the ability to capture epigenetic insights in humans, animals, plants and bacteria

Oxford Nanopore has released an all-context methylation model to enable real-time, high-accuracy epigenetic insights, integrating the capabilities of its flagship machine learning tool into the core basecaller and expanding its ability to detect methylation in non-human DNA, including plants and bacteria. This will accelerate research, enabling more comprehensive exploration of disease mechanisms, biomarkers, evolutionary dynamics, and the interplay between genes and the environment.

Announced during Oxford Nanopore’s annual London Calling customer conference, the new model was trained using synthetic DNA strands for the first time – a novel method that has the benefit of teaching the algorithm to see modified bases in a variety of different contexts. This will help to remove bias while expanding its ability to detect methylation in non-human DNA.

This update comes a year after Oxford Nanopore released “Remora,” a high-performance tool for methylation analysis that has since been fully integrated into operating software MinKNOW and production basecallers, broadening access to direct, PCR-free nanopore sequencing that captures methylation across the whole human genome. The process runs in parallel to standard basecalling and complements the simplicity of native DNA sample preparation, which can be done in just 10 minutes using the same run, at no additional cost.

With this release last year, nanopore sequencing became the most comprehensive and highest accurate technology for characterising methylation in native DNA, with the first releases aimed at targeting all CpG contexts. The most recent model is aimed at finding modifications in all contexts no matter where they occur, including 6mA.

Native sequencing for greater epigenetic insight

Methylation regulates the behaviour of cells. Although every cell in an organism has the same DNA, every cell has different functions. Methylation is integral to this process, working to turn different genes off or on. Scientists are now learning that sometimes this process goes awry, contributing to the cause of different diseases, including cancer.

Methylation detection has traditionally been dominated by using bisulphite treatment with short-read sequencing. Although this method has opened up the discovery of methylation sites, it also has limitations, increasing the cost and complexity of sequencing, owing to the requirement to repeat the run for essential comparison. Bisulphite cannot easily differentiate between methylation types such as 5mC/5hmC, or indeed add the detection of other modification types such as 6mA.

The Oxford Nanopore solution for detecting CpG and now all-context methylation provides a number of advantages. No additional, complex sample preparation is required and epigenetic modification analysis can be performed across the whole genome during the experiment. Therefore, no additional toxic chemistry is needed, and the phasing of base modifications is put in genomic context with the ability to sequence long fragments of DNA. Nanopore technology enables base-modification analysis to be performed alongside nucleotide sequencing on a single read basis, eliminating the need to sequence in high-depth.

When the original model was released last year it improved signal scaling, resulting in higher detection accuracy and quality filtered calls achieving 99.7% accuracy for 5mC in CpG contexts. This improved machine learning capability has enabled industry-leading methylation detection accuracy from a single read and provided biological insight into eukaryotic samples at only 20X coverage. Nanopore sequencing opens up the possibility of looking at methylation directly on individual DNA molecules, therefore representing how the DNA strands were in the cell, potentially the best and “most true” way of looking at DNA.

Marcus Stoiber, Principal Algorithms Researcher, Oxford Nanopore commented:

“Last year the release of our flagship machine learning model enabled the detection of methylation from raw nanopore signals at the highest accuracy, and its integration into our standard basecallers has led to groundbreaking research in human health and disease from the Nanopore Community. We are now rolling out an expanded set of models that will enable epigenetic insights in all living things, including plants and bacteria, for the first time. This expansion will allow for more immediate insights into critical areas such as bacterial infections, and we are excited to see what the research community does with this expanded tool.”

Methylation at London Calling 2023: At London Calling 2023, several speakers demonstrated the ability of nanopore technology to distinguish different DNA and RNA modifications in a range of different organisms. This research further demonstrates Oxford Nanopore is moving towards the detection of any modification, in any species, in any context. For highlights from London Calling, click here

Direct DNA and RNA sequencing for methylation analysis with nanopore sequencing (animation), click here

For a recap of the tech update at London Calling '23 delivered by Clive Brown, Chief Technology, Innovation and Product Officer, click here